Nvidia Continues to be The Market’s ‘Mammoth’, But Other Tech Giants And Startups Are Trying to Carve Out Their Niches

As Nvidia dominates its area of influence —controlling 80–90 percent of the AI training and inference accelerator market and $89 billion in data center revenue for Q2 2027 alone—startups and investors are carving out their own niches in an effort to push the market towards greater competition.

Amid these efforts, Nvidia maintains a strategic power, even against its biggest competitors like AMD. "Nvidia is not even an elephant. It's the mammoth in the room, and not for nothing," said Lior Prosor, co-founder and general partner at deep-tech venture fund Deep33. "They're doing everything that they can across the board to maintain that stronghold, and everyone else is trying to open up these little windows to bring in some sunlight where they can potentially win as well."

Even with Amazon's advancements with Trainium chips and Google's headway with its TPUs, Prosor said, "I don't see Nvidia dropping down dramatically in terms of market share, but the market itself is growing at such a pace that what's left of that, whatever 20 percent that is for Amazon, AMD and others, is worth more and more."

Despite Nvidia's authority, the more than $5 trillion company has itself heavily invested in AI infrastructure startups, including last year's $20 billion purchase of several assets from chipmaker Groq. Many of its investments, including in OpenAI, Intel and neocloud firms CoreWeave and Nebius Group, have circuitously come back to solidify its own footing by serving as its largest customers. Meanwhile, competitors challenging Nvidia with specialized custom capabilities, like Amazon's AWS and Google parent company Alphabet, remain core to Nvidia's own bottom line.

To challenge this monopoly, competitors realized they must work with the system, not against it. "Challengers win where Nvidia's biggest customers have already decided they need more sources," said Piotr Tomasik, co-founder and chief operating officer at AI neocloud firm TensorWave, which is powered exclusively by AMD GPUs. "OpenAI, Meta, and Anthropic made that decision in gigawatts. A challenger's job is to sit inside that demand, not create it alone."

The reality of the market is supply constraints aren't going anywhere. "As long as the market is as supply constrained as it's been, there is a buyer for that," said Prosor. "I don't see a future where we're not undersupplied."

But more competitors might not be the answer to an alleviated supply chain. "I don't think that more players are going to ease supply chain constraints for a number of years," said Tomasik. "The physical bottleneck only moves with capital at the foundry, packaging and memory layer."

Tomasik asserts players like Positron, Maverick Capital and his own Tensorwave—which is often posited a core challenger to Nvidia's lock on the ecosystem—are actively helping to diversify the industry.

Joe Fioti, co-founder and CEO at model compiler Luminal (part of startup accelerator Y Combinator's summer 2025 cohort), has come at the problem from a different lens. "Recreating that ecosystem isn't realistic for most companies, even AMD, because it's just too difficult," Fioti said. "A lot of companies are now taping out new chips and realizing that the software challenge is much harder than they anticipated. Many come from a hardware background and figure this out too late. They launch a new chip but can't sell it because nobody can program it. That's a very difficult problem to solve, especially as Nvidia continues to grow more powerful."

Luminal, which effectively serves as an ultra-efficient automatic translator for AI programs, works to improve inference speed and make chips interchangeable—potentially erasing Nvidia's advantage of locking in customers to its software-specific GPUs.

"We want to support as many chip startups and companies as possible because having more commercially viable chips means less of the extreme margin-taking we see in the industry today and a more commoditized compute market," said Fioti.

Nvidia's been at this for so long—and so successfully—that "everyone's thinking about how we create these vectors of innovation or non-Nvidia ecosystem areas," said Prosor. As the market expands with demand, it's only a matter of time until we see whether Nvidia retains its hyper-dominance or a more commoditized market begins to take root.