Editor’s note: In 2025, Crunchbase News heard from six active startup investors in artificial intelligence. Below, we publish highlights from those interviews or presentations. Read the full stories with Accel, Dell Technologies, Foundation Capital, GV, AI Fund and Sierra Ventures, as well as highlights from interviews in 2023 and 2024.
AI investment accelerated in 2025, as startups and investors alike angled for market share in this tech wave. By Q3, nearly half of all startup funding globally went to AI companies, according to Crunchbase data. Global venture funding overall was up 38% year over year in the third quarter, powered largely by megadeals for AI giants.
All told, AI startups raised around $100 billion in the first half of 2025 alone, roughly matching 2024’s full-year total.
Against that backdrop, six active AI investors shared their insights with us this year and offered a ground-level view of how the playbook is evolving, from compute and data moats to new models for co-founding companies. Here’s what we learned.
Accel’s Boterri: How startups can compete against behemoths
The “Super Six” companies — Nvidia, Microsoft, Apple, Alphabet, Amazon and Meta — generate hundreds of billions in operating cash flow, much of which is being poured straight back into AI infrastructure. Accel partner Philippe Botteri and his firm’s 2025 Globalscape report highlighted the new AI power map and how startups can compete.
The firm is one of the three most-active investors on the The Crunchbase Unicorn Board, which has surged in value this year amid the AI boom.
Accel in particular has backed a wave of native AI startups at both the model and application layer. Those startups include Anthropic, small model developer H Co., publicly traded AI infrastructure provider Nebius Group, and application-layer companies including Cursor-maker Anysphere, Perplexity, Synthesia and security startup Cyera.
While the incumbents are capturing enormous share, there’s still room for focused, fast-growing AI-native companies to wedge themselves into new categories or reinvent old ones, according to Boterri.
“If you don’t think that GenAI is going to generate a 1%-2% increase in the global GDP, then I’m not sure why we’re doing all this,” he said last month, speaking onstage at Web Summit in Lisbon, where Crunchbase News was also in attendance.
Where Foundation see opportunities in physical tech
But AI’s massive acceleration and computing needs also raise an existential question for the industry: Will the physical infrastructure keep up?
As many of the investors we spoke with this year noted, AI’s bottlenecks are increasingly physical — power, chips and data centers — and those bottlenecks are spawning some of the most interesting startup opportunities.
Botteri’s analysis points to an estimated 117-gigawatt energy shortfall over the next five years needed to serve projected AI demand — roughly equivalent to powering three large European economies combined.
It’s a problem that Steve Vassallo, general partner at Foundation Capital, is also coming at from the venture side. The firm incubated AI chipmaker Cerebras Systems back in 2016, well before today’s enthusiasm around AI infrastructure, and has gone on to back more than 100 AI startups.
In our interview, Vassallo recalled that betting on semiconductors in the mid-2010s “was a recipe for losing a lot of money” — but the team believed AI workloads were growing so fast that traditional architectures would eventually hit a wall. That thesis now looks prescient as Cerebras has signaled plans to go public, and chip giant Nvidia’s market cap has exploded to above $4 trillion.
Vassallo argued that the companies that matter most in this cycle will be those that both harness AI and are mindful of how humans can be “hacked” — in other words, products that respect psychology as much as physics. He pointed in particular to reinforcement learning with human feedback, where people help close the loop on AI behavior and, in the process, become more adept at working alongside the systems they’re training.
The firm’s AI investments include seed rounds into Tennr, a company automating authorization healthcare workflows from often convoluted and paper-based processes, and Jasper, an app that assists with writing built on top of OpenAI’s GPT-3. It also invested in the Series A for PlayerZero, whose product predicts and debugs software failures in AI-written code before it is deployed.
“We love working with founders who are living right at that edge,” Vassallo said.
Why Dell’s venture arm invests at the silicon level
At Dell Technologies Capital, or DTC, managing director Daniel Docter and partner Elana Lian sit at the intersection of infrastructure and enterprise demand.
Dell expects $20 billion in AI server shipments by fiscal 2026, and its venture arm has logged six exits since June — one IPO and five acquisitions — even as exits elsewhere in venture have been harder to come by.
As we discussed in our interview, Dell’s position as a leading GPU server provider means its venture arms sees nearly every serious enterprise AI buyer and builder up close.
And like other investors we spoke with, DTC noted that the current pace of investment in hot AI startups feels unprecedented. “We’ll meet with a company on a Tuesday for the first time and sometimes by Thursday, they have a term sheet that they’ve already signed,” Docter said.
The firm’s infrastructure-level investments include AI chipmaker Rivos, which Meta plans to acquire for an undisclosed amount. (The deal is pending regulatory approval.) It has also backed SiMa.ai, which makes a chip for embedded edge use cases including in automobile, drone and robot technologies, and Runpod, an AI developer software layer with on-demand access to GPUs.
DTC invests at the silicon level because you “can be incredibly disruptive to the ecosystem,” Docter told us.
At the application level, DTC’s investments include Maven AGI, which provides customer support for complex and high-compliance enterprise use cases, and Series Entertainment, a GenAI platform for game development that aims to drastically shorten deployment timelines.
Sierra Ventures’ layer-cake approach
If compute is the bottleneck, data is the differentiator.
Lian at DTC put it bluntly: “AI is almost a data problem.” For models to keep improving, she argued, you need high-quality, domain-specific data — not just more parameters.
At Sierra Ventures, managing partner Tim Guleri is also focused on data. As he explained in our interview, his firm tends to seek out startups that share a pattern: They attack big, painful workflows, promise order-of-magnitude productivity improvements, and sit on top of rich datasets.
Sierra’s “layered cake” framework breaks AI investing into five levels: infrastructure; applied infrastructure on top of foundational models; horizontal applications; vertical applications; and entirely novel innovations that wouldn’t exist without AI.
The firm isn’t trying to compete in the most capital-intensive infrastructure layer, but is leaning instead into applied infrastructure and applications where proprietary data and clever distribution can create durable moats, Guleri told us.
Global GDP is about $110 trillion, he noted, with roughly $6 trillion in agriculture — leaving more than $100 trillion in services and industries where he expects AI-driven efficiency gains to accrue.
AI is “the wave that’s lifting everything on top of it,” he said. “There’s going to be a tremendous amount of value creation in the coming decades.”
How a Google Brain co-founder builds and backs AI startups
Google Brain and Coursera co-founder Andrew Ng is taking a more hands-on route to unique data via corporate partnerships at AI Fund, his venture studio launched in 2018.
Corporate LPs including AES, HP, Mitsui & Co., Mitsubishi and others bring Ng and his AI Fund into highly specialized sectors — renewable energy, large-scale industrial operations, insurance and more — where internal data is both hard to access and critical to building defensible AI products.
Many of AI Fund’s startup ideas, Ng said, come directly from these partners spotting gaps in huge but under-digitized markets.
“It turns out a meaningful fraction of our startup ideas come from corporate partners that have spotted a market need, often in some sector of the economy, which is very large, very important but completely foreign to the typical consumer, or completely foreign to the typical AI engineer. I find that it’s been interesting how often we get to play in these spaces,” he said in our interview. “We think it’s wildly exciting, while no one else cares.”
Venture firms Sequoia Capital and New Enterprise Associates are also investors in the fund, but Ng said his fund’s strategy is differentiated from the traditional venture approach. “Unlike a traditional VC, our primary business activity is not to compete for deal flow,” he said. “Our primary business activity is to identify promising startup ideas, validate the market need and the customer need. Then we recruit a CEO to work alongside us to build a company.”
Ng said he sees continued opportunities in specific verticals such as visual and voice AI: “It feels like AI is not one thing; it is many different things that are creating new opportunities.”
GV on being willing to invest at AI’s premium valuations
Then there’s GV, which has quietly become one of the most-active and -flexible corporate investors in AI.
With Alphabet as its sole LP but independent on investment decisions, GV (formerly Google Ventures) has no qualms about backing startups that directly compete with Google’s own products — as it once did with Slack and is now doing with AI companies that go head-to-head with Alphabet’s internal initiatives.
Managing partners Dave Munichiello and Tom Hulme told us in our interview that they’re writing checks across the stack — from chips and compilers to applications — at both early and late stages.
And, they’re willing to accept premium AI valuations when they believe the opportunity warrants it.
“When we look at companies that are coming in to raise, the revenue run rate is insane. These companies are growing incredibly fast, faster than ever before,” Munichiello said. “And it’s very hard to spend a lot of time looking at AI applications companies, and then go back to looking at other companies.”
Related Crunchbase query:
Related reading:
Illustration: Dom Guzman
Stay up to date with recent funding rounds, acquisitions, and more with the
Crunchbase Daily.
