← The Wire·Semis·

OpenAI built inference chip in nine months with Broadcom help, outperforming Nvidia

OpenAI reportedly built an inference chip in nine months that outperformed Nvidia hardware, with Broadcom helping develop it. The report raises a competitive risk for Nvidia’s inference franchise while highlighting Broadcom’s custom-silicon opportunity.

NVIDIA’s headquarters, Santa Clara — file photoFile photo · NVIDIA’s headquarters, Santa Clara · Aug 2018 · Coolcaesar · CC BY-SA 4.0 · Source & license

The story

OpenAI developed an inference chip in nine months that beat Nvidia on inference performance, with Broadcom involved in making the project possible. The account establishes a strategic development effort without providing benchmark figures, production volumes, customer deployment details or a commercial launch timetable.

Nvidia enters the story from a position of substantial scale. Its fiscal-year revenue was $215.9B, up 65.5% YoY, with 71.1% gross margins and 55.6% net margins. Those figures show the size and profitability of the business potentially exposed to custom inference silicon, but they do not identify how much revenue depends specifically on inference workloads or how quickly customers could shift away from Nvidia systems.

Broadcom's reported fiscal-year revenue was $63.9B, up 23.9% YoY, with 67.8% gross margins and 36.2% net margins. Its role in the OpenAI project connects the company to the design and production ecosystem for custom accelerators, rather than positioning it as the owner of the competing AI model or chip platform. The involvement could support Broadcom's existing custom-silicon narrative, but the scale of the OpenAI work and whether it will become a material revenue line remain unclear.

The central claim also remains qualified by the lack of disclosed testing details. "Nvidia-beating" may refer to a specific inference benchmark, workload or efficiency metric rather than broad superiority across Nvidia's products. The degree to which OpenAI's chip is ready for large-scale deployment, whether it depends on Nvidia elsewhere in the data center, or whether Broadcom's contribution gives it an exclusive relationship is not established.

The next useful evidence would be a public benchmark, a production or deployment announcement, and disclosure of the commercial terms between OpenAI and Broadcom. Nvidia's future filings could also clarify the durability of AI infrastructure growth and the exposure of its revenue base to inference demand; Broadcom's filings or management commentary could show whether custom AI silicon is converting into material sales. Until those details appear, the development is a strategic signal with asymmetric implications for the competitive narrative rather than a quantified earnings revision.

Our read · Thu, Aug 27

OpenAI reportedly built an inference chip in nine months that outperformed NVDA hardware, with Broadcom (AVGO) helping develop it.

Why. The competitive implication is more credible than the immediate earnings impact: OpenAI’s reported development puts inference economics in focus, while Broadcom is positioned as an enabler of custom silicon. The absence of benchmark data, deployment scale and contract value keeps the read below conviction and leaves the relative benefit to AVGO unquantified.

Both cases

The case it holdsFor AVGO, the strongest hook is its reported role in helping OpenAI build custom inference silicon alongside $63.9B of existing revenue and 23.9% YoY growth.
The case it breaksFor NVDA, the report is not yet a demonstrated earnings threat because it gives no production scale, benchmark figures or evidence that customers are moving away from Nvidia’s $215.9B revenue base; for AVGO, the contract economics are undisclosed.

What could change this view: The report could describe a narrow benchmark or internal prototype that never reaches scaled deployment, leaving Nvidia’s broader platform and software position intact.

NVIDIA stock page →
Your side is graded privately against closes after 10 trading days. Research, not advice.

Who reported it · 1

Since this story · the companies named here, equal weight · −21.3%

Aug 27 · first close after publicationNov 19

Price context does not establish that the story caused the move.

Same arc

Reports linked to this one, newest first. Automatically linked reports may cover a broader event.

OpenAI Holds Back Astra Over Safety Risks, Iran Doubts US Deal Before MidtermsBloomberg Television · Tue, Sep 29 · 6:25 AM ET · automatically linked
OpenAI says governments among ‘dozens’ of organisations hacked by its agentsFinancial Times · Fri, Sep 25 · 6:18 PM ET · automatically linked
OpenAI expects to burn $280bn by 2030Financial Times · Fri, Sep 18 · 6:33 PM ET · automatically linked
Why Oracle Stock Jumped 6% Today on OpenAI Funding BuzzYahoo Finance · Thu, Sep 17 · 1:16 PM ET · automatically linked

Keep reading

More on NVDA

Alibaba aided China military-linked lab with AI computing technologyRansomware group The Gentlemen reportedly lists Nvidia for a leakDemocrats seek answers on Lutnick, Cantor and Nvidia lossesNvidia shares rise 3% after beat-and-raise quarter as JPMorgan flags volatilityNvidia chair rules out Huang return as CEO after 75% sales outlook

Latest on the Wire

Greer says US chip curbs will close China export-control loopholesBlackRock bitcoin ETF posts $541.1 million outflow as tech sell-off spreadsPoland says Russian missiles and drone crashed during Ukraine assaultAnthropic debut fails to halt technology-led slide in US stocksUS MATCH Act puts ASML at center of China chip export disputeContinue on the Wire →

Top this week

Nvidia allies lobby board to bring back Jensen Huang as CEOAnthropic Said to Guide IPO Pricing Above Market RangeNvidia interim CEO Colette Kress faces first earnings test after HuangAnthropic IPO could become the largest in US history on ThursdayBitcoin nears $64,000 as Anthropic IPO lifts Nasdaq futures nearly 2%
The Wire at 7 AM ET, free. Top stories with a bull and a bear case on each, every morning. No account needed.

About this report

Reported by Yahoo Finance as OpenAI Built an Nvidia-Beating Inference Chip in Nine Months. Broadcom (AVGO) Helped Make It Happen, . Who answers for this

The only report in this record. Closes are licensed end-of-day data from Databento; the move after a story is its first close against the close before it. How this is decided →

Kept as written · your side, if you take one, is graded privately against licensed closes after 10 trading days · nothing here is advice · How the Wire is made →