Nvidia’s supremacy in building computer chips for artificial intelligence (AI) has chilled venture funding for would-be rivals, investors said, with the number of US deals this quarter falling 80 per cent from a year ago.
The Santa Clara, California company dominates the market for chips that work with massive amounts of language data.
Generative AI models get incrementally smarter through exposure to more data, a process called training.
As Nvidia has grown stronger in this area, the harder it has become for companies attempting to build competing chips. Seeing these startups as a riskier bet, venture financiers are newly unwilling to provide big cash infusions.
Advancing a chip design to a working prototype can cost more than $500 million (€466 million), so the pullback has quickly threatened the startups’ prospects.
«Nvidia’s continued dominance has put a really fine point on how hard it is to break into this market,» said Greg Reichow, a partner at Eclipse Ventures.
«This has resulted in a pullback in investment into these companies, or at least into many of them».
Start-up deals and investment down
US chip start-ups have raised $881.4 million (€821.4 million) through the end of August, according to PitchBook data. That compares to $1.79 billion (€1.6 billion) for the first three quarters of 2022.
The number of deals has dropped from 23 to four through the end of August.
Nvidia declined to comment.
AI chip start-up Mythic, which has raised about $160 million (€149 million) in total, ran out of cash last year and was nearly forced to halt operations, technology website The Register reported.
But it managed to bring in a relatively modest $13 million (€12 million) investment several months later in March.
Nvidia has «indirectly» contributed to overall AI chip fundraising woes, because investors want «home run only type investments with a huge investment, huge return,» Mythic CEO Dave Rick said.
Difficult economic conditions have added to the downturn in the cyclical semiconductor industry, Rick said.
A secretive start-up called Rivos, which is working on chip designs for data servers has had trouble raising funding recently, said two sources familiar with the company’s situation.
A Rivos spokesperson said Nvidia’s market dominance hasn’t hindered its fundraising efforts and its hardware and software «continues to excite our investors».
Rivos is embroiled in litigation with Apple, which has accused Rivos of stealing intellectual property, and this has compounded the fundraising challenge.
Chip start-ups looking to raise cash are facing tougher demands from investors. They require companies to have a product that is within months of launch or already generating sales, sources said.
About two years ago, new investments in chip startups were often $200 million (€186.4 million) or $300 million (€280 million). That has fallen to about $100 million (€93 million), according to PitchBook analyst Brendan Burke.
At least two AI chip start-ups have overcome investor reluctance by trumpeting potential customers or their relationships with well-known executives.
Partner, Eclipse Ventures
To raise $100 million (€93 million) in August, Tenstorrent boasted about CEO Jim Keller, a near legendary chip architect who has designed chips for Apple, Advanced Micro Devices and Tesla.
D-Matrix, which has projected revenue of less than $10 million (€9.3 million) this year, raised $110 million (€102.5 million) last week, bolstered by financial backing from Microsoft and a commitment by the Windows maker to test d-Matrix’s new AI chip after it launches next year.
While these chip makers in Nvidia’s shadow struggle, startups in AI software and related technologies do not face the same constraints. They brought in about $24 billion (€22 billion) in funding this year through August, according to PitchBook data.
Despite Nvidia’s dominance in AI computing, the company does not have an unassailable lock on the sector. AMD plans to launch this year a chip that will compete with Nvidia’s, and Intel leapfrogged development by gaining a rival product in an acquisition. Sources see these as having long-term potential to become alternatives to Nvidia’s chip.
There are also adjacent applications that could provide openings for competitors. For example, chips that perform data-intensive computing for prediction algorithms are an emerging niche. Nvidia does not dominate this area and it’s ripe for investment.