Nvidia, Cerebras, and the OpenAI Trial: Tech Giants in the Spotlight
Nvidia's earnings report is expected amid the successful IPO of competitor Cerebras (+68%). The jury in Elon Musk's lawsuit against OpenAI also begins deliberations on the verdict.
Nvidia, Cerebras, and the OpenAI Trial: Cards Are Shuffled, Bets Are Placed
Three news items that flashed across the feed in the last 48 hours may look like scattered events from the world of tech giants to the average reader. But those inside the industry see a different picture — this is not a coincidence, but perfectly orchestrated timing. Nvidia's earnings, Cerebras' IPO, and the verdict in Musk v. OpenAI are three points of a single triangle that will define the AI market architecture for the next two years. Here's what the media is missing.
The Gist: What's Really Happening
On May 20, Nvidia will report for the first quarter of fiscal 2027. Consensus is $79.2 billion in revenue, growth of nearly 80% year-over-year. These are colossal numbers, but the market is watching two things: confirmation of the Vera Rubin timeline and the revenue forecast for the next quarter. Analysts expect guidance around $90 billion.
Against this backdrop, on May 14, Cerebras went public on the NASDAQ — a company that marketing headlines have dubbed the "Nvidia killer." The IPO priced at $185 per share at a $488 billion valuation; shares now trade at $303.5. Eight months ago, the company was worth $81 billion — a sixfold increase in less than a year.
And on May 18, two days before Nvidia's report, the federal court in Oakland put an end to Elon Musk's lawsuit against OpenAI. The jury deliberated for less than two hours. Verdict: Musk filed too late; the three-year statute of limitations had expired. The path to an OpenAI IPO worth up to $1 trillion is open.
Three events in five days. A coincidence? No. Synchronization that reshapes the balance of power.
Timeline and Context
If you haven't been following the developments, here's a condensed sequence leading to the current moment:
- 2015 — Musk and Altman create OpenAI as a nonprofit. Mission: "AGI for the benefit of humanity."
- 2024 — Media reveals OpenAI's plans for a massive restructuring into a commercial entity.
- October 2025 — OpenAI completes its transition to a Public Benefit Corporation (PBC). The nonprofit retains 26%, Microsoft gets about 27%.
- April 2026 — The trial begins on Musk's lawsuit. He demands the restructuring be overturned, Altman and Brockman removed, and damages exceeding $130 billion.
- May 14, 2026 — Cerebras conducts the largest AI IPO of the year, raising $55.5 billion.
- May 18, 2026 — The court rules in favor of OpenAI.
- May 20, 2026 — Nvidia publishes its report, which will either confirm the continuation of the AI supercycle or trigger a wave of profit-taking.
Who Wins and Who Loses
Sam Altman wins. The court decision is not just the removal of legal risk. It's a green light for the IPO, scheduled for Q4 2026. Target valuation: up to $1 trillion, expected raise around $60 billion. For comparison, that's more than the sum of all major tech IPOs over the past decade. OpenAI generates about $2 billion in monthly revenue, and weekly active ChatGPT users exceed 900 million. Musk has announced an appeal, but an appeal won't have the same paralyzing effect as an active lawsuit threatening licenses and contracts.
Jensen Huang wins — but not as obviously as it seems. The market expects strong numbers from Nvidia. Blackwell accounts for 70% of high-end GPU shipments; Rubin enters production in June. But there's a nuance: 65% of Nvidia's data center revenue comes from inference. And in inference, GPUs are most vulnerable. The B200 at batch=1 has over 99% tensor core idle time. The "memory wall" problem hasn't gone away: over 10 years, computation has accelerated 1000x, memory bandwidth only 10x. Hence the strategic move — the $20 billion acquisition of Groq last year. Huang knows the vulnerability but doesn't talk about it publicly.
Cerebras wins — but with caveats. The WSE-3 shows 18x speed over the B200 on Llama 3.3 70B. That's a real advantage for the narrow scenario of "one query, fast answer." But the company's S-1 contains alarming numbers: 86% of 2025 revenue came from two UAE clients — MBZUAI and G42. The US accounted for only 14%. OpenAI signed a $10+ billion contract, but the same OpenAI also inked a deal with AMD for 6 GW of capacity. No major player is betting exclusively on Cerebras — it's a diversification option, not the main route.
Elon Musk loses — tactically. His xAI is rushing to go public: a merger with SpaceX, ticker SPCX, date June 12, target valuation $1.75 trillion. But now his main enemy Altman has a free hand, and Musk loses leverage. Making loud statements after a court defeat is harder.
What the Media Isn't Saying
Non-obvious insight: Cerebras' real goal is not to compete with Nvidia, but to be acquired by Nvidia or AMD within 18–24 months. All the talk about a "GPU killer" is a narrative for the IPO, not a business strategy. Look at history: Groq was bought by Nvidia, Graphcore by SoftBank, UntetherAI by AMD. Cerebras is the only one that went public instead of M&A. Why? Because a $488 billion market cap gives currency for acquisition, not competition. The WSE-3 is a point solution for the decode phase of inference. No CUDA ecosystem, no training stack, no prefill chips. Cerebras isn't building a full platform — it's building an asset for sale.
Second hidden story: The 44 GB of SRAM on the WSE-3 is both an advantage and a fatal limiter. The entire architecture is optimized for a fast response to a single query. But the main trend of 2026 is AI agents that need long dialogues with growing context. Stratechery analyzed: as long as data fits in 44 GB, Cerebras flies; once it exceeds, the advantage disappears. OpenAI knows this, hence diversifying purchases.
Third point — the trial and OpenAI's IPO are no coincidence. Altman's legal team deliberately dragged out procedural stages to get a court ruling 5–6 months before the planned IPO. Now they have a "clean legal history" for the SEC. Musk fell into a trap: his own tweets from 2020 became evidence that he knew about OpenAI's commercialization long before filing the lawsuit.
Forecast: Next 30 Days and 90 Days
May–June 2026 (30 days):
Nvidia will publish a strong report and guidance above $85 billion for the next quarter. That's enough for a 5–7% stock rise, but not enough for a new rally — the market is already pricing in the "best case." If guidance comes in below $80 billion, a 10–15% correction is possible, dragging the entire sector down. Cerebras will likely correct more sharply — investors will start counting the real client base and notice the dependence on two counterparties.
On June 12, Musk's SpaceX/xAI IPO. It will be the largest listing in history and will siphon liquidity from other AI assets. Nvidia and Cerebras may dip on this news.
June–August 2026 (90 days):
Cerebras will announce its first major partnership outside the UAE — either with a US hyperscaler or a European telecom. This is needed to diversify the client base, something the S-1 is silent about.
OpenAI will begin its IPO roadshow. The valuation will be lowered from $1 trillion to the $750–850 billion range — investment bankers will realize the market isn't ready to pay a 40x revenue premium for a structure with a nonprofit controlling shareholder.
Key signal I'm watching: Will Nvidia announce the acquisition of another inference startup before the end of summer? If yes, it means Huang sees the threat from Cerebras as more serious than he publicly admits. If no, it means the internal Rubin roadmap truly covers inference needs, and Cerebras loses its window of opportunity.
Bets are placed. The next 90 days will show who was playing the long game and who was bluffing.
— Editorial Team
No comments yet.