• Microsoft is diversifying the AI models for 365 Copilot to reduce reliance on OpenAI, per Reuters.
• The move aims to cut costs and improve speed for Microsoft's enterprise clients.
• Big Tech firms have invested heavily in AI startups to develop advanced models.

Microsoft is diversifying the artificial intelligence models it uses to power its flagship AI assistant, 365 Copilot, in a bid to reduce its dependence on OpenAI, Reuters first reported.

The Big Tech giant is moving toward adding internal and third-party AI models to help run its 365 Copilot to cut costs and address concerns about speed for its enterprise clients, per the report.

As one of OpenAI's main backers and corporate partners, Microsoft will continue working with the AI startup to develop frontier models.

"We incorporate various models from OpenAI and Microsoft depending on the product and experience," Microsoft said in a statement to Reuters.

Microsoft can customize OpenAI's model, per its original licensing agreement with the company. While Microsoft is currently training its own model, Phi-4, the software juggernaut is looking at modifying other third-party models to reduce the cost of running 365 Copilot.

Microsoft's lackluster debut of Copilot raised concerns about the software giant's ability to deliver on its AI ambitions, BI previously reported. Some customers appear to be dissatisfied with the product, spurring complaints that it is ineffective, expensive, and not secure.

In the race to develop powerful frontier models, Big Tech giants have scrambled to bulk up their arsenal of AI investments — and pumped billions of dollars into startups to help them achieve this goal.

Amazon has invested $8 billion into AI juggernaut Anthropic and used the startup's technology to power its digital assistant. This year, Google signed a deal with Character.AI, a startup that develops anthropomorphic chatbots, which allowed it to hire its founder and license its technology — a deal described as an "acquihire."

Microsoft and OpenAI did not immediately respond to a request for comment, sent outside standard working hours.

After three years of tense reductions, the skies are clearing over Silicon Valley, and startup investors seem broadly optimistic about a resurgence in tech dealmaking.

We asked venture capitalists at 35 firms like Andreessen Horowitz, Insight Partners, IVP, and Sapphire Ventures, to tell us what's hot and what's not in tech next year, how potential regulatory changes could rouse a sleepy exit market, and where artificial intelligence goes from here.

In 2025, venture capitalists expect a loosening of antitrust regulations under the new presidential administration. This could reignite acquisition activity by strategic buyers, which would allow funds to distribute proceeds from those deals to their own investors, or limited partners, and raise new funds to invest in the next generation of startups, said Brian Garrett, managing director at Crosscut Ventures.

In recent years, startups weren't the only ones facing a cash crunch. Established funds raised the lion's share of funding dollars, while many newish and boutique funds struggled to raise. A torrent of dealmaking, combined with Trump's return to the White House and an end to the political uncertainty, could mobilize investors in these funds who had been sitting on the sidelines to whip out their checkbooks, said Ivan Nikkhoo, a managing partner at Navigate Ventures.

"Uncertainty breeds defense, optimism breeds offense," said Matt Murphy, a partner at Menlo Ventures and early Anthropic investor. "We're going into a cycle where acquirers are feeling they need to play offense and startups feel like it's time to invest in leadership. And the IPO market is open for best-in-class assets."

From IPOs to robotaxis, these are the tech trends to watch in 2025, according to venture capitalists.

Infrastructure cools off, apps soar

Jai Das, president and partner at Sapphire Ventures: "A larger number of 'application layer' companies will have a breakout year with several crossing $100 million in revenues. I predict 50 companies will cross $50 million ARR while still growing 60%+, and at least 10 will hit $100 million ARR. A lot of these companies will be prosumer companies, but there will be several business application companies as well."

Ben Lerer, managing partner at Lerer Hippeau: "When you get the cost of compute going down as quickly as it has, and the number of options in terms of foundational models growing as it has, you end up with a really interesting time for the application layer to thrive. If you're a startup, you can go with the flavor of the month — not just a ChatGPT wrapper, or a Claude wrapper, or a Gemini wrapper, or you name it — but some combination of all of them to optimize functionality, results, and the cost of those results."

Lower rates kick the IPO market into gear

Sofia Dolfe, partner at Index Ventures: "2025 is the year we will see the IPO market opening back up. There are already signs that this is on the horizon: we're seeing gradual recovery, rates have started to come down, and there are many later-stage companies with the financial profiles to go public."

Michael Yang, senior managing partner at Omers: "Two kinds of companies will go public as the IPO window opens back up next year. First, the truly great businesses that are really scaled and have forecastable growth and would've gone public earlier if the IPO market was more favorable, and second, companies that entered into structured financings with dirtier terms that need to go public for timing reasons."

Nima Wedlake, managing director at Thomvest Ventures: "The IPO market will remain closed for most tech companies, with a high bar for entry — $300 million-plus ARR, fast growth, and cash-flow breakeven or better."

As crypto prices surge, founders return to the drawing table

Nihal Mehta, general partner at Eniac: "Guidance on what the regulations could be for crypto and AI would encourage founders to build productively within those areas."

Jai Das, president and partner at Sapphire Ventures: "The new administration is crypto-friendly, bringing with it an expected acceleration of crypto-based business models (especially those using stablecoins). I predict we'll have another crypto mania in 2025."

Some venture funds go belly-up

Wesley Chan, cofounder and managing partner at FPV Ventures: "In 2025, I predict a lot of contraction for VCs, except for top funds. We're still in a downturn. Some firms shut down, a lot of firms are not doing new deals, and you will see a lot of junior-mid level employees leave."

The great funding bifurcation continues

Molly Alter, partner at Northzone: "The 'sexiest' deals will continue to raise at sky-high valuations, but for the rest of the pack, companies will need to show very specific metrics to command a strong valuation. There will be a great bifurcation into the 'haves' and the 'have-nots.'"

Don Butler, managing director at Thomvest Ventures: "Startup shutdowns will increase, particularly at the seed stage, as companies run out of cash. This will influence valuations, with investors likely focusing on startups that have shown resilience or achieved meaningful milestones."

Matt Murphy, partner at Menlo Ventures: "Valuations will rise as growth rates and market multiples recover, but many companies still might not grow back into their ZIRP valuations. People are over that and won't let it get in the way of pursuing opportunity. Valuations for GenAI companies will continue to be outliers based on any historical metrics."

Robotaxis cover new terrain

Brian Walsh, head of Wind Ventures: "2025 will be the year that we enter the age of 'robo taxis' with, first, Waymo now well along its adoption S-curve in San Francisco and expanding quickly, and, second, Tesla favorably positioned with quickly maturing best-in-class autonomy technology (no human in the loop) and an existing large fleet to scale it."

Kasper Sage, managing partner at BMW i Ventures: "Autonomous fleet deployments will gain traction in controlled, high-density environments such as for applications like campus environments and logistics for heavy industries."

Trump policy heralds return of megadeals

Aaron Jacobson, partner at NEA: "With the change of administration, I expect the return of mega M&A deals. We are going to see a 'WhatsApp' like $20 billion-plus M&A outcome for a leading AI company."

Michael Yang, senior managing partner at Omers: "Big Tech will be back at the M&A table with a new administration and regulatory regime in place. They've been quieter in recent times but should be chomping at the bit to capitalize on what is still a buyer's market."

Funding rounds become even more fluid

Sasha McKenzie and Van Jones, both deal leads at Wellington Access Ventures at Wellington Management: "The concept of letter rounds in VC is becoming more amorphous. We're seeing $30 million and $100 million seed rounds, raising questions about what seed even means anymore. The model is shifting towards evaluating how quickly founders can run and how disciplined they are with results, rather than hitting historically stated milestones (e.g., $1 million in revenue to raise a Series A). There will be more nuance in how VCs evaluate progress, focusing more on the operator and their ability to balance vision with execution, based on the capital they have."

Multi-agent systems take center stage

Aaron Jacobson, partner at NEA: "Chatbots are overhyped. Agents are under-hyped. Enterprises will move beyond the low-hanging fruit of 'GPT-wrappers' to deploy digital workers that can reason and take action to make a real business impact."

Praveen Akkiraju, managing director at Insight Partners: "If 2024 was the year of LLMs, we believe 2025 will be the year of agentic AI — where highly capable state-of-the-art reasoning LLMs are combined with orchestration frameworks like memory, tool calling, and user-in-the-loop processes to build AI agents that can address progressively complex business workflows."

Seema Amble, partner at Andreessen Horowitz: "In the short term, human workers will be the reviewer in the loop; in the future, as trust is established over time, I expect many data-derived actions will shift toward being entirely a set of narrowly defined task-driven agents."

S. Somasegar, managing director at Madrona: "The world where we each have a digital assistant that works with a collection of AI agents is probably five to ten years out. But having AI agents that can do specific tasks really, really well is happening sooner and I think we will see a ton of progress on this in 2025."

Tender offers grow for a selective group of companies

Ravi Viswanathan, founder and managing partner at NewView Capital: "The venture secondaries market will continue to be an important source of liquidity — a trend we think is here to stay due to structural dynamics of the venture asset class."

Simon Wu, partner at Cathay Innovation: "The size of tender offers has grown from millions to billions as the desire to own top-performing names by mutual funds and VCs increases, thus allowing some of the best names to stay private longer. Tenders are likely to get bigger to a selective group of companies in tandem with a more active IPO market next year."

Industry-specific software takes over

Molly Alter, partner at Northzone: "Vertical SaaS will become more highly valued than ever, due to the increasing difficulty of differentiating a product in horizontal categories."

Cathy Gao, partner at Sapphire Ventures: "Vertical software will evolve rapidly as AI moves to the agentic phase, enabling end-to-end automation of complex, industry-specific workflows that were once beyond the reach of software. By pairing deep domain expertise with intelligent automation, vertical AI will unlock new use cases, deliver outsized ROI, and become table stakes for staying competitive."

Fintech roars back

Alexa von Tobel, managing partner at Inspired Capital: "Given the new political climate, we, of course, expect to see less regulation across the board. I think we'll see acceleration in a few core categories, including fintech."

Marlon Nichols, managing partner at MaC Venture Capital: "Fintech is an area I'm excited to invest in, particularly fintech startups leveraging AI to create transformative personal finance tools."

Sydney Thomas, general partner at Symphonic Capital: "We are watching the regulatory environment towards fintech ease which has enabled massive speculation on what asset class will win. … This also means, many startups will be required to regulate themselves, which isn't always an easy thing to do."

Robots join society

Claire Yun, investor at Piva Capital: "Generative AI will continue to accelerate and supercharge robotics; simultaneously, we will see a choke point in human labor as an aging domestic workforce and protectionist policies create a sharp supply and demand imbalance. The result will be a colorful Cambrian explosion of robots as they step in to fill this gap."

Bob Ma, partner at Wind Ventures: "Urban areas will have fleets of robots on sidewalks, while drones will manage suburban and rural deliveries. Enhanced speed, cost-efficiency, and sustainability will redefine retail and e-commerce, with regulations supporting wider adoption and innovation."

Yuri Lee, partner at IVP: "As AI advances enable robots to move from structured, repetitive tasks to more complex and dynamic real-world applications, we'll see rapid progress in robotic perception, manipulation, and decision-making capabilities."

Small language models rise in popularity

Tasneem Dohadwala, partner at Excelestar Ventures: "Small language domain-specific models are starting to show more value. Instead of using vast swaths of the internet to train large models, these smaller models can be trained on specific datasets, such as medical journals, newspapers, or email collections. As a result, they are highly tailored and more accurate in reflecting a user's particular constraints and voice.

Michael Yang, senior managing partner at Omers: "If 2024 was the year of the LLMs, 2025 will be the year of small language models (SLMs) and proprietary data sets spawning the next generation of enterprise SaaS applications. Companies have realized that data in their midst can be harnessed in new and better ways than the 'structured workflow apps' of old and by leveraging targeted SLMs, they can do work differently, more efficiently."

Founders flock to private equity

Brad Bernstein, managing partner at FTB Capital: "Despite the IPO market showing better performance in Q3'24 with proceeds already surpassing 2023 totals, structural issues like regulatory burdens and governance challenges still pose obstacles for small and mid-cap companies. Private equity markets are stepping in to fill the gap, with growth equity deals comprising a larger share of activity and providing opportunities for startups in high-growth sectors like insurtech and healthcare tech."

Jai Das, president and partner at Sapphire Partners: "With the new administration, I predict we will see an uptick in exits, and much more tech M&A activity. We'll also see PE firms buying up a lot of companies once boards and management teams realize these businesses won't be able to grow at 30% at scale and ultimately, IPO."

Open-source foundation models come for OpenAI and xAI's lunch

Aaron Jacobson, partner at NEA: "Open-source foundation models will close the gap with the leading proprietary models. On top of this we will see a significant shift away from pre-training models from scratch to fine tuning OSS models and distilling them to smaller models for faster performance."

Mo Jomaa, partner at CapitalG: "I predict that in 2025 we will continue to see open source technologies consume the infrastructure layer in software. We have seen this trend play out in several categories already, including data and analytics (which led to our investment in Databricks) and observability (which drove our investment in Grafana). Enterprises will continue to adopt open source because it helps them save money, avoid vendor lock-in, and shape the product roadmaps of the technologies that they procure."

Record deals and dollars flow to cyber and national security

Andrew Schoen, partner at NEA: "We will see a surge of investment into technologies critical to restarting the US industrial base and enhancing national security. A record number of deals and dollars will go into AI, automation, cybersecurity, and frontier technology serving manufacturing, supply chain, and defense markets."

Jake Seid, general partner at Ballistic Ventures: "Over the next 18 months, we're going to see a lot more cybersecurity exits. While this may include an uptick in M&A activity, I expect we'll see cybersecurity companies go public in 2025 and in the first half of 2026 given how large the market for cyber products has become."

Trump's tech advisors bend his ear

Samir Kumar, general partner at Touring Capital: "We should expect a lot less regulatory headwinds in 2025 for AI given David Sacks will be the AI and crypto czar for the new administration. This is likely to even result in the repeal of President Biden's executive order on AI."

Francesco Ricciuti, associate at Runa Capital: "In the US, Trump is bringing prominent people from the startup and VC world in the government, and I wouldn't be surprised if the regulatory landscape will evolve towards entrepreneurship and technology."

• Women are gaining ground in venture capital, with more becoming partners in 2024.
• In 2023, women held 19% of investment partner roles in US funds, up from 16% in 2020.
• Firms like Andreessen Horowitz and IVP are promoting women to partner positions globally.

A new generation of women tech investors is kicking down the doors of venture capital's boys' club. In 2024, VC heavyweights such as Andreessen Horowitz, Insight Partners, and IVP, elevated women investors to partner positions.

On European soil, names like Cherry Ventures, AlbionVC, and Ascension championed this transformation across the continent.

Despite facing a tougher path to promotion in recent years, a number of women rose to the role of partner or general partner for the first time in 2024.

According to a 2023 report by PitchBook, women made up 19% of investment partners at US funds, up from 16% in 2020. Similarly, across the Atlantic, a report by European Women in VC revealed that women held 16% of general partner roles in Europe in 2023.

Business Insider is recognizing the women in VC who became partners or general partners in 2024, both in Europe and the US. While this list celebrates those within established organizations, it offers a testament to many other women forging their paths with micro funds and angel investing, setting a vibrant precedent for the future.

Cat McDonald took the mantle of partner at UK fund AlbionVC, where she will lead generalist investments.

Grace Ge joined Amplify as a partner from Menlo Ventures. She invests in AI-enabled software, developer tooling, data, and infrastructure.

Andreessen Horowitz promoted longtime investor Jennifer Li to general partner to help manage a $1.25 billion infrastructure fund.

Iulia Tudor was promoted to partner at UK-based firm Ascension, where she supports its fund focused on energy, fintech, and health.

Prerna Sharma, an early Uber employee, was promoted to general partner at Antler VC. She's based in Texas.

Nadine Torbey was elevated to partner at AlbionVC. She will help lead the fund's deep tech, AI, and enterprise software investments.

Asymmetric Capital Partners promoted Nancy Chou to partner to invest in software, fintech, and insurtech.

Rachel Wilson moved to The BMF Fund as a managing partner to continue backing Black and diverse founders.

Dinika Mahtani was promoted to partner at Cherry Ventures, having led investments in Swap, Finesse, and Julienne Bruno.

Dr. Carolin Althoff became a partner at Cusp Capital, where she backs digital startups in the environmental, social, and governance domains.

Lily Wang made partner at Expon Capital. She has been sourcing key investments for the Luxembourg-based firm and covers sectors such as energy, AI, and health.

Foreground Capital promoted cofounder and women's health investor Alice Zheng to partner.

Claire Zau made partner at GSV Ventures, where she spearheads its AI investments.

Lavanya Bhamidipati was promoted to partner at InHealth Ventures, a London-based, healthcare-focused fund.

Initialized Partners elevated investor Zoe Perret to partner. She invests in AI, energy, and climate.

Insight Partners promoted infrastructure and cybersecurity investor Philine Huizing to partner.

Inspired Capital elevated Charlotte Ross to partner after five years with the firm.

Yuri Lee made partner at IVP after sourcing deals like medical scribe startup Abridge and healthcare provider Accompany Health.

Esha Vatsa joined London-based seed-stage fund Mercuri as a partner.

Ex-Uber employee Vivian Cheng became a partner at Next47, investing in AI and software.

Erika Nash became a partner at Next Frontier Capital, after holding roles at Pelion Venture Partners and OpenView.

Cecilia Zhao joined General Catalyst's London office as a partner. She focuses on fintech and enterprise software.

Charlotte Salasky was promoted to partner at Notion Capital, where she recently helped to fundraise its €300 million, fifth fund.

Rebecka Löthman Rydå joined Norrsken Accelerator as a general partner. She brings 10 years of experience as an active VC investor.

Molly Alter made partner at multistage firm Northzone, after holding roles at Insight Partners and Index Ventures.

Tara Stokes rose to the role of partner at Point72 Ventures and leads the firm's AI and machine learning investments.

Plus VC promoted Zainab Al Sharif to partner to invest in early-stage startups across the Middle East and North Africa.

Louise Kingston was promoted to partner at the venture capital arm of wealth management firm Baird, where she also leads its internal data projects.

QED Investors promoted Camila Vieira to partner. She leads its Brazil investments.

Megan Kelly made partner at Threshold Ventures, where she works closely with portfolio companies Odaia and QA Wolf.

VamosVentures promoted Ashley Ryder to partner to continue backing Latinx and diverse founders.

Lily Bernicker made partner at Wireframe Ventures, where she invests in climate and health.

Vera Baker is a partner at French VC 4P Capital, leading the firm's early-stage investment strategy.

• Tiun has secured a $2.5 million pre-seed funding round to innovate media monetization.
• The startup offers a pay-for-what-you-use model to reduce friction for media consumers.
• Business Insider got an exclusive look at the 9-slide deck it used to raise fresh funding.

Zurich-based startup Tiun has raised $2.5 million in pre-seed funding for its digital wallet, which lets users pay only for the online media they consume and provides businesses with an alternative way to monetize their content.

The startup aims to streamline the online subscription process, which its cofounder and CCO Nikolaos Christoforakos says is "full of friction" — from registering and authorizing an account to committing to an upfront subscription.

"Our mission is to help media providers — and more broadly service providers — to attract, engage, and convert younger audiences," Christoforakos told Business Insider.

Instead of subscribing to multiple individual media providers, Tiun offers users a digital wallet they can use across multiple media partners to purchase preferred content on a "pay-for-what-you-use" basis.

The account is free to create, with users topping up funds using an established payment method. This eliminates the need to enter new payment details for every outlet and eases the friction in the consumer journey, Christoforakos said.

Tiun gets a slice of each transaction, splitting the rest with the media provider. Companies providing the content — from podcasts to streaming to online news — can also access Tiun's business suite with metrics on reader data and conversions.

"There's a value to the product outside of the infrastructure, and on top of that, we make money with a revenue share," Christoforakos told BI.

Media organizations have been exploring micropayment models for several years — but with limited success. Dutch online news platform Blendle, one of the main champions of micropayments, later ditched its model in favor of subscriptions. Christoforakos says that Tiun is not positioning itself as a micropayment service that's in competition with previous efforts.

But the startup is providing the infrastructure that addresses one of the oft-cited reasons for the micropayment model's slow adoption: the lack of a standard payment method across media outlets.

"Our vision is to redefine how the next generation will consume and pay for online content by establishing a new standard that improves interoperability between wallets and applications," Christoforakos said.

Tiun's funding round was led by Swiss VC firm Founderful, with additional funds coming from Blue Wire Capital and a16z scout Maximilian Lehmann, among other angel investors.

With the fresh funding, Tiun will develop some core product offerings in the coming year.

We got an exclusive look at the 9-slide pitch deck used to secure the fresh funding.

• Cloud security startup Wiz has grown rapidly in just four years, raising $1.9 billion along the way.
• The AI boom has accelerated cloud adoption, says Raaz Herzberg, the CMO of Wiz.
• Wiz rejected a $23 billion Google acquisition offer in July and said it plans to IPO instead.

In early 2020, Raaz Herzberg was a product manager at Microsoft Azure when she was offered a position at Wiz, a newly created cloud security startup. Since then, Wiz has turned down a $23 billion takeover offer from Google, expanded into Europe, and reached $500 million in annual recurring revenue.

Its next priority is to double that revenue metric — and then become a public company.

Wiz was launched by four cofounders who sold their previous business, Adallom, to Microsoft. It bills itself as a cloud security company that helps companies identify risks in their cloud providers.

"It felt like an opportunity I couldn't possibly pass on," Herzberg, now the chief marketing officer and vice president of product strategy at Wiz, told Business Insider in an interview from the company's new London office. "I started as head of product — but we started in the worst time. This was March 2020, when COVID-19 started," she said. "It's easy to remember because everything changed immediately."

It turned out that the pandemic was a boon for business. As more companies shifted to remote work, they increasingly relied on cloud services — expanding Wiz's client base. Four years since its launch, the scaleup has raised over $1.9 billion in funding from investing heavyweights such as Andreessen Horowitz, Thrive Capital, and Index Ventures.

It now sees huge opportunities to use the AI boom to cement its position in the market before it launches its initial public offering.

AI adoption has supercharged Wiz's business

Cloud computing offers crucial infrastructure underpinning AI applications. As more companies rush to adopt AI, security and privacy have taken center stage.

"The adoption of AI is very similar to what happened with the accelerated use of cloud," Herzberg said. Wiz has found that over 80% of its customers are using AI services — which "are, in some ways, like cloud services. Companies don't buy their machines or chips, so they're using these technologies on the cloud," she added.

"Part of the reason we're growing so fast is because we have access to the public cloud, which is growing incredibly fast — and AI only pushes that growth further," she said.

Because AI services are often used on public clouds — a service offered by third parties rather than an internal network — cloud security has become a critical issue.

Wiz's rapid growth has also been bolstered by its arsenal of acquisitions. This year, it scooped up security startups Rafft, Gem, and Dazz in a bid to bulk up its engineering talent and product suite.

"When we acquire companies, we don't sell their product. They rebuild it from scratch in the Wiz infrastructure," Herzberg told BI, noting that the company is still on the hunt for more acquisitions. "We believe in this concept of growing inorganically."

European expansion on the road to IPO

Wiz's global headquarters is in New York, with offices in Virginia, Texas, Colorado, and Israel. In August 2024, Wiz established its European headquarters in London. Its plush office is a short walk from Silicon Roundabout — London's scaled-down answer to Silicon Valley.

"The European market has been an ideal fit for our technology because it's more constrained by security, and more privacy aware than the US market," Herzberg told BI. "We estimate we will be able to get 35% of our revenue from Europe."

Wiz works with industry heavyweights on the continent, such as Revolut, Tide, and BMW. In 2024, it says it reached $500 million in ARR — but is aiming to reach $1 billion before it IPOs.

Operating independently is a big priority for the company. Earlier this year, Wiz turned down a $23 billion offer from Google, instead opting to prime itself for a public debut.

Herzberg declined to comment about the Google deal but added that Wiz had lofty ambitions to establish itself as a market leader in the cloud security domain.

"We are building a company that I believe can be the biggest cybersecurity company in the world," she said. "And I think at this point we are on that path."

She likened companies buying security services to buying insurance packages, pointing to incumbents in the security industry with a similar model.

"So if I look at the wing of, for example, network security firewalls, it has a clear leader — people today buy Palo Alto firewalls. They used to buy checkpoint firewalls. Now the leader is Palo Alto," she said. "Another example would be like the endpoint production server protection, that's a big domain on its own, and then it has a clear leader, like CrowdStrike."

As the public cloud domain balloons, Herzberg believed there needs to be a "leader in place to protect that domain" — adding that Wiz had a goal of taking that mantle.

Growth ambitions

A slate of elections this year has pushed the company to prepare for more government-mandated cybersecurity measures, especially in the US. In anticipation of Donald Trump's second term, Wiz has started building on a federal sales strategy.

Still, Herzberg said the scaleup is in "no rush" to go public. It's now looking for a chief finance officer, a requirement for companies that want to IPO.

"With the place we are at in terms of revenue and publicity and everything, it just brings us better candidates than we had as little as a year ago," Herzberg said. "I don't think we necessarily need someone with cloud security experience," she added. "We're hoping to announce that hire soon."

Elsewhere, it's gearing up to release two new products as it cements its presence in Europe with an impending hiring spree.

Still, the team hopes to come full circle to its New York roots by the time it's ready to IPO.

"Which one has the gong?" Herzberg laughed. "New York is where we'd list."

• Nscale has raised $155 million in Series A funding for its hyperscaler platform.
• The startup offers everything from access to data centers to GPU infrastructure.
• BI got an exclusive look at the pitch memo the startup used to secure the fresh funds.

Nscale, a London-based startup providing companies with access to data centers and clusters of AI chips, has raised $155 million in fresh funding.

The startup, which came out of stealth in May 2024, bills itself as a fully integrated AI infrastructure platform.

As a hyperscaler, Nscale provides the "full stack" of technologies companies need to train and run AI applications like large language models. That includes data centers, software, and graphics processing units, Nscale's founder and CEO Joshua Payne told Business Insider in an interview.

The startup differentiates itself from AI cloud providers, such as Lambda Labs and Coreweave, which offer only specific components, such as GPUs, in the AI infrastructure layer. By providing everything from its own data centers to virtualized GPU nodes, Payne said that the company could leverage better unit economics than its competitors.

"The problem for the industry is chicken and egg. Take an LLM customer — they may want 10,000 GPUs, but they don't have the expertise for that," he told BI. "Many of our competitors don't own their own data centers, so they have to license them. In our case, we have all of that in-house. Given that we are able to own all those segments in the value chain, we're faster and cheaper."

Payne said that Nscale pivoted to focus on AI infrastructure across the full stack after the public release of ChatGPT-3. "So our thesis was, if this is indeed the fourth industrial revolution as people are claiming, then how would you build a resilient AI cloud that would survive commodity cycles? said Payne. "We found the best way to do that was to vertically integrate it, building both data centers, GPUs, and software."

Since it emerged from stealth, it has rapidly grown its pipeline of greenfield data center capacity across Europe and the US from 300 megawatts to 1.3 gigawatts.

The startup makes its money by building data centers, purchasing GPUs, and deploying AI cloud services, which it then leases on an hourly basis. Clients can sign contracts for any given period of time to use the services.

Sandton Capital Partners led the $155 million round, which also included participation from Kestrel, Bluesky Asset Management, and Florence Capital. The startup previously raised more than $30 million in pre-seed and seed funding, with the size of those early-stage rounds reflecting investors' heightened appetite to back startups operating in the AI infrastructure layer.

With the fresh funding, Nscale plans to invest in large clusters of GPUs and also double down on software development for its public cloud platform, which will be released in January 2025.

BI got an exclusive look at the 9-slide memo the startup used to secure fresh funding.

• The US has introduced new export controls on China's semiconductor industry, targeting 140 firms.
• They aim to curb China's AI and defense-tech growth, partly because of national security concerns.
• Trump has called China the "main threat" to the US AI industry.

The US announced on Monday another round of export controls on China's semiconductor industry, multiple outlets reported, weeks ahead of Donald Trump's second term.

Washington plans to restrict exports to 140 Chinese companies, including the chip-equipment heavyweight Naura Technology Group, to curb China's growing capabilities in artificial intelligence and defense technology.

It is the third crackdown the US has initiated on China's chipmaking industry since October 2022. The move is set to also stop the export of advanced high-bandwidth memory, a key component in the development of AI chips in China.

The chip manufacturers Semiconductor Manufacturing International Corp. and Huawei are also on the list of 140 companies.

"They're the strongest controls ever enacted by the US to degrade the People's Republic of China's ability to make the most advanced chips that they're using in their military modernization," Gina Raimondo, the US's commerce secretary, told reporters from the Financial Times, The New York Times, and others on Sunday.

The US has long been embroiled in a technological race against China, which developing AI and military tech at pace.

The Biden administration's latest sanctions are partly driven by national security concerns that China's access to high-quality chips could allow it to bolster its military applications, especially through the use of AI.

Last month, Reuters reported that researchers in China affiliated with the People's Liberation Army had used Meta's open-source AI model Llama to develop an AI tool that could be applied to military use cases.

While this latest wave of measures is from the Biden administration, Beijing reportedly anticipates further sanctions from Trump. China has been attempting to stockpile chips from the US in recent months, with purchases reaching $1.11 billion in October, an analysis of customs data from the South China Morning Post found.

Previous comments from Trump suggest he's aligned with the Biden administration when it comes to thwarting China's AI growth. In a June interview with the social-media personality Logan Paul, Trump billed China as the "main threat" to the US AI industry. "We have to be at the forefront," he said.

There are still some internal disagreements about the approach to restricting Huawei chip-production facilities, the Financial Times said. Some of the Chinese tech giant's chip-production plants were not included on the list, with one person close to the discussions telling the FT that they were not in operation, so it's not clear whether they would be used for the production of advanced chips.

Mao Ning, a spokesperson for China's foreign ministry, said at a press conference earlier this month, "China is firmly opposed to the US overstretching the concept of national security, abusing export control measures, and making malicious attempts to block and suppress China."

The US Department of Commerce didn't immediately respond to a request for comment from Business Insider made outside normal working hours.

• TSMC founder Morris Chang asked Nvidia founder Jensen Huang to take over as CEO in 2013.
• In a new memoir, Chang reveals he set out his vision for TSMC for 10 minutes before Huang declined.
• Huang said, "I already have a job," Chang recalled.

The founder of Taiwanese chip giant TSMC has revealed he once asked Jensen Huang if he would succeed him as the company's CEO.

But Huang, the founder and CEO of AI chipmaker Nvidia, turned the role down in less than 10 minutes and said "I already have a job," Morris Chang wrote in his memoir, published Friday.

In the memoir, Chang writes he was looking for a successor to lead TSMC in 2013. He said that Huang's character, professional background, and deep knowledge of the semiconductor space made him an ideal frontrunner for the role.

Huang listened intently as Chang spent 10 minutes explaining his ambitions for TSMC, but said he was determined to keep his focus on Nvidia, Chang writes.

Nvidia has since become one of the world's most valuable companies, fuelled by the AI boom. Huang has been CEO and president since founding it in 1993.

Huang and Chang have shared an amiable relationship that spans their professional endeavors.

In the early years after its launch, Nvidia exclusively partnered with TSMC to produce its chips. In 1998, TSMC helped supply Nvidia with production workers when it was short-staffed.

Nvidia now works with various chipmakers but remains one of TSMC's biggest customers, along with Apple.

Having founded TSMC in 1987, Chang, 93, stood down as CEO in 2018 and was replaced by C C. Wei, the current CEO. According to Forbes, Chang has a personal wealth of $4.1 billion.

The latest memoir, his second volume of autobiography, details his life from 1964 to 2018.

• Amazon is developing a new AI model called Olympus for video analysis, The Information reported.
• The AI tool could help search for specific scenes in video archives.
• Amazon-backed Anthropic has already launched its own multimodal model for images and videos.

Amazon may have doubled down on its investment in Anthropic, but it also appears to be hedging its AI bets by developing its own model that can process images and video in addition to text.

The tech giant's new model, code-named Olympus, could help customers search video archives for specific scenes, The Information reported.

It's a type of AI — known as multimodal — already offered by Anthropic, the startup that Amazon pumped a fresh $4 billion into earlier this month, bringing its total investment into the company to $8 billion.

Amazon could launch Olympus as soon as next week at its annual AWS re:Invent conference, The Information reported.

Amazon's partnership with Anthropic goes beyond capital. The e-commerce juggernaut has used Anthropic's technology to power its digital assistant and AI coding products. And Amazon Web Service customers get early access to a key Anthropic feature: fine-tuning their data through Anthropic's chatbot Claude.

In return for Amazon's most recent investment in Anthropic, the startup said it would use AWS as its "primary cloud and training partner." The deal also includes an agreement for the OpenAI rival to use more of Amazon's chips.

The development and launch of Olympus could reduce Amazon's dependency on Anthropic for multimodal AI, especially if the new video model becomes as a cheaper alternative.

The Big Tech giant has a huge repository of video archives, which could be used to train the AI model for various use cases, from sports analysis to geological inspections for oil and gas companies, per the report.

Amazon doesn't have a seat on Anthropic's board, but it takes a portion of the proceeds from its sales, as its platform runs on AWS servers.

Amazon and Anthropic did not immediately respond to a Business Insider request for comment.

• PyannoteAI is in talks to raise $10 million in initial funding, sources told Business Insider.
• The startup's AI platform detects and identifies speakers in audio transcriptions.
• Voice AI is gaining traction, with startups like ElevenLabs in talks to raise significant funding.

A French startup developing voice intelligence models is in talks to raise around $10 million in initial funding, Business Insider has learned from two people with knowledge of the deal.

PyannoteAI, which launched in 2024, has developed a model to automatically detect and identify speakers in audio transcriptions. The concept, known as "speaker diarization," aims to enhance audio transcriptions so that they can be better deployed in industries such as customer service and sales.

The startup is in talks to raise funding at a valuation of around $40 million, the two sources said.

Details of the latest funding round are not yet finalized, and the figures involved are subject to change.

Pyannote's cofounder and CEO, Vincent Molina, confirmed to BI that the company is "in the process of fundraising" but said he could not provide further details.

According to its website, the startup's platform is based on research conducted by Hervé Bredin, its other cofounder. Bredin has a research background in speaker diarization and deep learning for audio, speech, and natural language processing.

France has established itself as a hub for AI development since 2022, with buzzy upstarts such as Mistral and H raking in hefty early-stage funding just weeks after their launch. According to Dealroom, French startups that bill themselves to be in the generative AI space have raised over $2.29 billion this year, surpassing any European country.

AI tools for generating voices have also gained traction this year. Voice-cloning startup ElevenLabs is looking to raise funding at a $3 billion valuation, Business Insider first reported. This year, OpenAI released Voice Mode, a tool that allows users to have spoken conversations with ChatGPT.

• The rate of AI model improvement appears to be slowing, but some tech leaders say there is no wall.
• It's prompted a debate over how companies can overcome AI bottlenecks.
• Business Insider spoke to 12 people at the forefront of the AI boom to find out the path forward.

Silicon Valley leaders all-in on the artificial intelligence boom have a message for critics: their technology has not hit a wall.

A fierce debate over whether improvements in AI models have hit their limit has taken hold in recent weeks, forcing several CEOs to respond. OpenAI boss Sam Altman was among the first to speak out, posting on X this month that "there is no wall."

Dario Amodei, CEO of rival firm Anthropic, and Jensen Huang, the CEO of Nvidia, have also disputed reports that AI progress has slowed. Others, including Marc Andreessen, say AI models aren't getting noticeably better and are all converging to perform at roughly similar levels.

This is a trillion-dollar question for the tech industry. If tried-and-tested AI model training methods are providing diminishing returns, it could undermine the core reason for an unprecedented investment cycle that's funding new startups, products, and data centers — and even rekindling idled nuclear power plants.

Business Insider spoke to 12 people at the forefront of the AI industry, including startup founders, investors, and current and former insiders at Google DeepMind and OpenAI, about the challenges and opportunities ahead in the quest for superintelligent AI.

Together, they said that tapping into new types of data, building reasoning into systems, and creating smaller but more specialized models are some of the ways to keep the wheels of AI progress turning.

The pre-training dilemma

Researchers point to two key blocks that companies may encounter in an early phase of AI development, known as pre-training. The first is access to computing power. More specifically, this means getting hold of specialist chips called GPUs. It's a market dominated by Santa Clara-based chip giant Nvidia, which has battled with supply constraints in the face of nonstop demand.

"If you have $50 million to spend on GPUs but you're on the bottom of Nvidia's list — we don't have enough kimchi to throw at this, and it will take time," said Henri Tilloy, partner at French VC firm Singular.

There is another supply problem, too: training data. AI companies have run into limits on the quantity of public data they can secure to feed into their large language models, or LLMs, in pre-training.

This phase involves training an LLM on a vast corpus of data, typically scraped from the internet, and then processed by GPUs. That information is then broken down into "tokens," which form the fundamental units of data processed by a model.

While throwing more data and GPUs at a model has reliably produced smarter models year after year, companies have been exhausting the supply of publicly available data on the internet. Research firm Epoch AI predicts usable textual data could be squeezed dry by 2028.

"The internet is only so large," Matthew Zeiler, founder and CEO of Clarifai, told BI.

Multimodal and private data

Eric Landau, cofounder and CEO of data startup Encord, said that this is where other data sources will offer a path forward in the scramble to overcome the bottleneck in public data.

One example is multimodal data, which involves feeding AI systems visual and audio sources of information, such as photos or podcast recordings. "That's one part of the picture," Landau said. "Just adding more modalities of data." AI labs have already started using multimodal data as a tool, but Landau says it remains "very underutilized."

Sharon Zhou, cofounder and CEO of LLM platform Lamini, sees another vastly untapped area: private data. Companies have been securing licensing agreements with publishers to gain access to their vast troves of information. OpenAI, for instance, has struck partnerships with organizations such as Vox Media and Stack Overflow, a Q&A platform for developers, to bring copyrighted data into their models.

"We are not even close to using all of the private data in the world to supplement the data we need for pre-training," Zhou said. "From work with our enterprise and even startup customers, there's a lot more signal in that data that is very useful for these models to capture."

A data quality problem

A great deal of research effort is now focused on enhancing the quality of data that an LLM is trained on rather than just the quantity. Researchers could previously afford to be "pretty lazy about the data" in pre-training, Zhou said, by just chucking as much as possible at a model to see what stuck. "This isn't totally true anymore," she said.

One solution that companies are exploring is synthetic data, an artificial form of data generated by AI.

According to Daniele Panfilo, CEO of startup Aindo AI, synthetic data can be a "powerful tool to improve data quality," as it can "help researchers construct datasets that meet their exact information needs." This is particularly useful in a phase of AI development known as post-training, where techniques such as fine-tuning can be used to give a pre-trained model a smaller dataset that has been carefully crafted with specific domain expertise, such as law or medicine.

One former employee at Google DeepMind, the search giant's AI lab, told BI that "Gemini has shifted its strategy" from going bigger to more efficient. "I think they've realized that it is actually very expensive to serve such large models, and it is better to specialize them for various tasks through better post-training," the former employee said.

In theory, synthetic data offers a useful way to hone a model's knowledge and make it smaller and more efficient. In practice, there's no full consensus on how effective synthetic data can be in making models smarter.

"What we discovered this year with our synthetic data, called Cosmopedia, is that it can help for some things, but it's not the silver bullet that's going to solve our data problem," Thomas Wolf, cofounder and chief science officer at open-source platform Hugging Face, told BI.

Jonathan Frankle, the chief AI scientist at Databricks, said there's no "free lunch " when it comes to synthetic data and emphasized the need for human oversight. "If you don't have any human insight, and you don't have any process of filtering and choosing which synthetic data is most relevant, then all the model is doing is reproducing its own behavior because that's what the model is intended to do," he said.

Concerns around synthetic data came to a head after a paper published in July in the journal Nature said there was a risk of "model collapse" with "indiscriminate use" of synthetic data. The message was to tread carefully.

Building a reasoning machine

For some, simply focusing on the training portion won't cut it.

Former OpenAI chief scientist and Safe Superintelligence cofounder Ilya Sutskever told Reuters this month that results from scaling models in pre-training had plateaued and that "everyone is looking for the next thing."

That "next thing" looks to be reasoning. Industry attention has increasingly turned to an area of AI known as inference, which focuses on the ability of a trained model to respond to queries and information it might not have seen before with reasoning capabilities.

At Microsoft's Ignite event this month, the company's CEO Satya Nadella said that instead of seeing so-called AI scaling laws hit a wall, he was seeing the emergence of a new paradigm for "test-time compute," which is when a model has the ability to take longer to respond to more complex prompts from users. Nadella pointed to a new "think harder" feature for Copilot — Microsoft's AI agent — which boosts test time to "solve even harder problems."

Aymeric Zhuo, cofounder and CEO of AI startup Agemo, said that AI reasoning "has been an active area of research," particularly as "the industry faces a data wall." He told BI that improving reasoning requires increasing test-time or inference-time compute.

Typically, the longer a model takes to process a dataset, the more accurate the outcomes it generates. Right now, models are being queried in milliseconds. "It doesn't quite make sense," Sivesh Sukumar, an investor at investment firm Balderton, told BI. "If you think about how the human brain works, even the smartest people take time to come up with solutions to problems."

In September, OpenAI released a new model, o1, which tries to "think" about an issue before responding. One OpenAI employee, who asked not to be named, told BI that "reasoning from first principles" is not the forte of LLMs as they work based on "a statistical probability of which words come next," but if we "want them to think and solve novel problem areas, they have to reason."

Noam Brown, a researcher at OpenAI, thinks the impact of a model with greater reasoning capabilities can be extraordinary. "It turned out that having a bot think for just 20 seconds in a hand of poker got the same boosting performance as scaling up the model by 100,000x and training it for 100,000 times longer," he said during a talk at TED AI last month.

Google and OpenAI did not respond to a request for comment from Business Insider.

The AI boom meets its tipping point

These efforts give researchers reasons to remain hopeful, even if current signs point to a slower rate of performance leaps. As a separate former DeepMind employee who worked on Gemini told BI, people are constantly "trying to find all sorts of different kinds of improvements."

That said, the industry may need to adjust to a slower pace of improvement.

"I just think we went through this crazy period of the models getting better really fast, like, a year or two ago. It's never been like that before," the former DeepMind employee told BI. "I don't think the rate of improvement has been as fast this year, but I don't think that's like some slowdown."

Lamini's Zhou echoed this point. Scaling laws — an observation that AI models improve with size, more data, and greater computing power —work on a logarithmic scale rather than a linear one, she said. In other words, think of AI advances as a curve rather than a straight upward line on a graph. That makes development far more expensive "than we'd expect for the next substantive step in this technology," Zhou said.

She added: "That's why I think our expectations are just not going to be met at the timeline we want, but also why we'll be more surprised by capabilities when they do appear."

Companies will also need to consider how much more expensive it will be to create the next versions of their highly prized models. According to Anthropic's Amodei, a training run in the future could one day cost $100 billion. These costs include GPUs, energy needs, and data processing.

Whether investors and customers are willing to wait around longer for the superintelligence they've been promised remains to be seen. Issues with Microsoft's Copilot, for instance, are leading some customers to wonder if the much-hyped tool is worth the money.

For now, AI leaders maintain that there are plenty of levers to pull — from new data sources to a focus on inference — to ensure models continue improving. Investors and customers just might have to be prepared for them to come at a slower pace compared to the breakneck pace set by OpenAI when it launched ChatGPT two years ago.

Bigger problems lie ahead if they don't.

• ElevenLabs is raising new funding, boosting its valuation to over $3 billion, according to multiple sources.
• The startup offers tools for creating voice agents that mimic the user's own voice.
• ElevenLabs is well-funded, but OpenAI's secretive voice project could pose a major threat.

ElevenLabs, a startup that can recreate a person's voice from a recording, is close to raising a new round of funding this year that would push its valuation to over $3 billion, Business Insider has learned from two people with knowledge of the deal.

This time, a year ago, the voice-cloning startup was worth $100 million. The startup is aiming to raise $200 million in the round, according to one source.

The potential round is being led by the wealth-management juggernaut Iconiq Capital, both people said, and included participation from the venture capital firm Andreessen Horowitz, one of the people said.

Details of the latest funding round are not yet finalized, and the figures involved are subject to change.

The story of ElevenLabs began with founders Piotr Dąbkowski and Mati Staniszewski's desire to improve voice-dubbing in movies. Their idea turned into an advanced system for turning text into speech that can now create original dialogue bearing a striking resemblance to the user's own voice. The tool only needs minutes of audio of someone speaking to clone their voice.

The company, which sells mostly to publishers, content creators, and media and entertainment companies, has reached $90 million in annual recurring revenue, according to the two people familiar with the new funding. It's pacing to hit $100 million in annual revenue by year's end.

That kind of growth had investors flinging term sheets at the two-year-old company, the people familiar with the deal said. The company has been increasing its targeted round size and valuation during the process of raising funding, according to these sources.

Funding for voice technology companies has picked up even as some of the shine comes off the digital intelligence category as a whole. ElevenLabs competes with Vapi, a voice-based developer tool with $2 million in seed money from Kleiner Perkins and Abstract Ventures, and companies like Thoughtly and Retell that are developing voice agents for call centers.

ElevenLabs has gathered more money than any of them, though the real competition may come from one of the world's most valuable startups. In March, OpenAI, the maker of ChatGPT and Dall-E, shared details about a pilot program that will allow a closed group of companies to create a synthetic voice based on a real person's speech.

Business Insider emailed ElevenLab's co-founders on Wednesday morning and did not receive a response. Representatives Iconiq Capital, and Andreessen Horowitz did not immediately respond to requests for comment.