company
Publicly traded chipmaker (NASDAQ: CBRS) whose wafer-scale silicon with 44 GB on-die SRAM powers OpenAI's Ultrafast tier; published the GDP-Val and Humanity's Last Exam speed claims.
No evidence-backed relationships are recorded.
OpenAI has contracted 750MW of Cerebras wafer-scale inference capacity, delivered in three 250MW segments by the end of 2028. The dates are contractual, but neither company has published pricing, latency figures or a plan for which customers get access.
Reality
Cerebras says splitting inference stages across chip types gave 5x more throughput from the same number of its systems without slowing token generation. Because the count covers only Cerebras hardware, the figure does not yet show what a mixed-chip fleet costs per unit of work.
Reality
OpenAI's Ultrafast tier runs GPT-5.6 Sol up to 14 times faster than standard, at as many as 750 output tokens a second, for a limited preview group. With no price published, teams can prototype real-time features on it but cannot yet budget a production launch.
Reality
Gimlet Labs plans 100 MW of Cerebras-powered inference capacity, pairing wafer-scale chips with GPUs for different phases of each request. For buyers, the quoted 3,000 tokens per second is a company target until measured results appear.
Reality
Bloomberg reports the London startup raised $100 million in seed money to route each task to the cheapest workable model-chip pairing. The wager is that heterogeneity outlasts convenience.
Reality
The accelerator is in full production and the headline number is a single-request generation rate at 100,000 tokens of context. That is a different purchase order than throughput.
Perspective Coverage
3 publishersReality
Jalapeno's lead is measured against last generation and the volumes are tiny, but a first-pass ASIC clearing Nvidia, AMD and Google parts reprices the design barrier, not the supply chain.
Perspective Coverage
7 publishersReality
DigitalOcean says step-by-step reasoning is billed as output and invisible by design. The 90% figure it cites comes from a paper that estimates hidden token counts, so moving it onto your own invoice takes a matching task mix.
Reality
In IEEE Spectrum's account of the 2026 compute market, reasoning and agentic workloads have pushed serving toward memory-heavy silicon, and Amazon now runs a single inference job across two vendors' chips.
Reality
Positron says the money covers a TSMC N3P tapeout by the end of 2026, a new 2MW engineering data center and a production ramp in the second half of 2027. The larger of the round's two tranches is described as up to $500 million.
Reality
A dev.to harness put five coding tasks through four APIs and every run passed its verifier on the first attempt, so the only thing left to compare is clock time, where one trial per cell sets a fragile order.
Reality
The gate on recursive self-improvement is how long one loop takes, and experiments, training runs and new chips all take real time. That makes capability curves fast but bounded, which changes what you sequence first.
Reality
More memory on a Cerebras wafer arrives with CS-6, two generations out. Everything shipping before then, Nexus included, is rack engineering around the SRAM budget the wafer already has, and long contexts are where that budget hurts.
Reality
Forge counts seven of twenty VC-backed debuts trading above their IPO price. Figma, up 250% on day one, is 84.3% below its first-day close.
Reality
Forge's data puts today's AI cohort at $100 billion in about five years, against 16 for the SpaceX generation. The listing increasingly looks like an exit for someone else.
Reality
The first third-party benchmark of the LP30 rack came in at roughly four times the next-fastest public endpoint, measured one request at a time on a model small enough to fit.
Reality
The Nexus rack, not the WSE-3, is now the thing Cerebras ships. That makes upgrade cadence the number buyers should price, and 10,000 tokens per second a target rather than a plan input.
Reality
Morgan Stanley closed its purchase of the secondaries marketplace in January, and the private shares changing hands there increasingly sit behind lenders who get paid first.
Reality
The Information reports a valuation more than 50% above the last round, with a licensing arrangement considered alongside it. Buyers of AI search should read the roadmap as supplier-driven.
Reality
Hardware from the largest acquisition in Nvidia's history is in production and headed to a neocloud this year. The market sold it anyway, two days before earnings.
Reality
Invest · August 20, 2026 · 3 publishers
Invest · August 20, 2026 · 3 publishers
Product · August 19, 2026 · 2 publishers
Product · August 17, 2026 · 1 publisher
Product · August 17, 2026 · 1 publisher
Security · August 15, 2026 · 1 publisher
Product · August 15, 2026 · 1 publisher
Build · August 15, 2026 · 1 publisher
Product · August 15, 2026 · 1 publisher
Product · August 14, 2026 · 2 publishers
Build · August 14, 2026 · 3 publishers
Invest · August 14, 2026 · 3 publishers