Beyond GPUs, a new global startup war is erupting around inference, memory, interconnects, and CPUs. Why put it that way? Recent fundraising numbers in AI chips show that the sector is entering an almost manic phase.
On August 18, U.S. AI chip startup Etched announced a new $700 million round at a $21 billion valuation. Less than a month earlier, on July 23, Etched had just closed a $300 million Series C at a $10.3 billion valuation. In barely a month, the company’s valuation doubled.
Almost on the same day, a far less well-known AI chip company, Velaura AI, announced a $110 million Series A that pushed its valuation past $1 billion, making it a unicorn overnight.
Pull the timeline back a bit, and the same story has already played out again and again this year.
In February, Cerebras raised another $1 billion at a valuation of about $23 billion—nearly a fourfold jump from $8.1 billion four months earlier. In March, South Korea’s Rebellions raised $400 million at a valuation of about $2.34 billion. In May, the U.K.’s Fractile closed a $220 million Series B. Canada’s Taalas raised $169 million in February, bringing total funding to about $219 million—only to be acquired by AMD in August, less than six months later. Tensordyne is preparing a Series D later this year; its new inference system already has more than $200 million in projected demand.
These are not isolated fundraising events. Semiconductor Engineering data show that in Q1 2026 alone, 80 semiconductor startups raised a combined $8.4 billion, and 18 of them closed individual rounds of more than $100 million. More striking is where the money is going: a large number of companies are not building another traditional GPU. They are betting on AI inference, or trying to solve bandwidth and data-movement problems at the chip, memory, interconnect, or even full data-center level. Photonics has therefore remained a hot investment theme.
After more than a decade of “survival of the fittest,” AI chip entrepreneurship suddenly looks back. But this time, the game is completely different.
