AI hardware startup Etched is reviewing incoming investment bids ranging from $40 billion to $50 billion, according to people familiar with the matter. This rapid re-pricing occurs just a couple of months after Etched secured $700 million at a $21 billion valuation in October 2026, following a $300 million round led by Sequoia at a $10.3 billion valuation in July. Such back-to-back funding tranches with escalating valuations have become the default playbook for the market's buzziest infrastructure plays.
Building proprietary silicon is an exceptionally capital-intensive endeavor. If Etched closes at the top of these incoming bids, the cash injection will buy roughly 3.5 years of operational runway, as noted by sources familiar with the offers. The four-year-old company now operates a 10-megawatt data center in Silicon Valley and established a facility in Taiwan to park itself right next to TSMC's manufacturing bottlenecks.
Inference-Driven Architecture and Market Demand
Investor enthusiasm centers entirely on the company's architectural bet: optimizing exclusively for inference rather than general-purpose training. Etched designed two new components from scratch to accelerate inference, which co-founder and COO Robert Wachen defines as the compute that happens after a user hits enter. The company claims its ASICs can process tokens faster and cheaper than Nvidia's dominant GPU lineup.
Etched has designed two new components from scratch to speed up inference — the computing process that happens after a user submits a prompt.
Quantitative trading firm Jane Street led the previous $700 million round and already took delivery of an early system, demonstrating immediate enterprise adoption where performance latency dictates profit. For high-frequency shops like Jane Street, a microscopic advantage in execution speed translates directly into massive returns. Etched reported $1 billion in pre-orders following the test chip run at TSMC over the summer.
Talent Acquisition and Competitive Positioning
To unseat an incumbent, you need their engineers. Etched has poached roughly 15% of its 400-person workforce directly from Nvidia, according to The Wall Street Journal, systematically acquiring the silicon design talent required to execute its roadmap.
This $40 billion price tag is a clear market signal: capital is abandoning general-purpose infrastructure in favor of hyper-specialized inference hardware. Anyone allocating next quarter's compute budget needs to audit their vendor roadmaps immediately — paying for universal GPUs when your workload is 90% inference is an expensive habit.