Five developments in AI public market valuation, legal liability, enterprise deployment, infrastructure opposition, and semiconductor supply risk.

Cerebras Systems (ticker: CBRS) is an AI chip company whose core product, the Wafer Scale Engine 3, is built by treating an entire silicon wafer as a single processor rather than cutting it into individual chips – an approach the company says runs AI inference, the process of generating outputs from an already-trained model, up to 15 times faster than competing products. The company priced 30 million Class A shares at $185 on May 13, 2026, and began trading on the Nasdaq Global Select Market on May 14. The stock opened at $350 – nearly double the offer price – reached an intraday high of $385, and closed at $311.07, a 68% first-day gain. The offering raised $5.55 billion. On May 15, shares fell a further 10.08%, closing at $279.72.

Cerebras had originally marketed shares at $150–$160, raising both the size and the price range in response to demand that exceeded available shares by more than 20 times. Revenue at the company grew 76% in 2025 to $510 million, and the company generated net income of $88 million, compared with a net loss of $481.6 million the prior year. One material risk disclosed in regulatory filings: more than 85% of Cerebras' 2024 revenue came from G42, a UAE-based AI company, a concentration that previously drew US regulatory scrutiny before being cleared. Cerebras has also established partnerships with Amazon and OpenAI, with OpenAI launching its first AI model running on Cerebras chips earlier in 2026. The valuation at Day 1 close was ~$70 billion based on outstanding shares per regulatory filings, or ~$86 billion on a fully diluted basis per Bloomberg data; Reuters placed the fully diluted figure at $56.43 billion at the IPO price – a range reflecting genuinely different methodologies for counting diluted shares.

Closing arguments in Musk v. Altman concluded on May 14, 2026, at the US District Court in Oakland, California, before Judge Yvonne Gonzalez Rogers. The nine-person jury – comprising six women and three men – will begin deliberations on Monday, May 18. Their verdict is advisory only: the judge retains final authority on both liability and remedies. Simultaneously, a separate remedies phase begins on May 18, in which Judge Gonzalez Rogers alone will hear arguments on potential damages and structural relief. Elon Musk was absent from closing arguments; his attorney cited his presence on President Trump's Beijing delegation alongside Tim Cook, Jensen Huang, and Larry Fink.

The case turns on two claims: that OpenAI's 2025 recapitalisation – which converted its original nonprofit structure into a capped-profit company – breached the charitable trust under which Musk made ~$38 million in donations between 2015 and 2017; and that OpenAI CEO Sam Altman, co-founder Greg Brockman, and Microsoft were unjustly enriched in the process. Microsoft, named as a co-defendant on an aiding-and-abetting theory, has invested a cumulative $13 billion directly in OpenAI; CEO Satya Nadella testified that Microsoft's total spend including infrastructure exceeded $100 billion. Musk's legal team has asked the court to unwind the 2025 restructuring, remove Altman and Brockman from their roles, and disgorge up to $134 billion – none of which would go to Musk personally. A threshold question on the statute of limitations is determinative: Judge Gonzalez Rogers wrote last month that a finding against Musk on that question would "highly likely" lead her to direct a verdict for the defendants. The most pointed piece of evidence on Musk's side: co-founder Greg Brockman's 2017 personal journal, introduced at trial, described OpenAI's nonprofit framing as "a lie."

OpenAI announced the launch of the OpenAI Deployment Company, referred to internally as DeployCo, on May 11, 2026. The venture is backed by more than $4 billion in initial investment from 19 global partners and is majority-owned and controlled by OpenAI. Lead investors are TPG, Advent, Bain Capital, and Brookfield; other named partners include Goldman Sachs, SoftBank Corp., Warburg Pincus, McKinsey & Company, Bain & Company, and Capgemini. The venture is valued at $10 billion. OpenAI has guaranteed its private equity backers a minimum annual return of 17.5% over five years. Simultaneously, OpenAI announced plans to acquire Tomoro, an applied AI consulting firm whose clients include Mattel, Red Bull, Tesco, and Virgin Atlantic; the Tomoro acquisition is expected to add ~150 Forward Deployed Engineers – specialists who work directly at client sites – subject to regulatory approval.

DeployCo's model is explicitly patterned on Palantir's embedded-engineering approach: its engineers will work inside client organizations to identify high-impact use cases, connect OpenAI models to internal data and business processes, and build production-ready AI systems. The partners' portfolio companies – collectively spanning more than 2,000 businesses – form DeployCo's initial customer base. Anthropic announced a $1.5 billion parallel enterprise joint venture backed by Blackstone, Hellman & Friedman, and Goldman Sachs in the days before the DeployCo launch, also focused on embedding AI into enterprise workflows. OpenAI's Chief of Applications Fidji Simo told staff in a company-wide meeting that Anthropic's enterprise gains should serve as a "wake-up call." Shares of Infosys, Tata Consultancy Services, and HCLTech fell sharply on news of the DeployCo launch, as investors assessed the risk that embedded AI engineering from model providers could displace demand for conventional implementation services.

A Gallup survey of 1,000 US adults conducted March 2–18, 2026 – the first time Gallup has tracked this question – found that 71% of Americans oppose building an AI data center in their local area, including 48% who are strongly opposed. Only 27% support local data center development, with just 7% strongly in favor. Environmental concerns dominate: 46% of respondents said they worry "a great deal" about the environmental impact of AI data centers, and another 24% worry "a fair amount." Opposition to AI data centers now exceeds local opposition to nuclear power plants (53% opposed in the same March survey), even though nuclear opposition has not exceeded 63% since Gallup first tracked that question in 2001.

The Gallup findings land against a backdrop of massive ongoing construction. ConstructConnect's May 2026 Data Center Report shows $46.5 billion in data center starts through March 2026, against a comparable 2025 year-to-date total of just $7.3 billion – a more than sixfold pace increase. ConstructConnect also identified a $63.2 billion late-stage pipeline for the remainder of 2026, with ~$28 billion in late-stage projects in Texas, Oklahoma, and Arkansas and another ~$22 billion along the Eastern Seaboard. The 12-month moving average puts 2026 spending on pace for ~$121 billion. In a follow-up April web survey, half of data center opponents cited excessive resource consumption, with water usage and power grid pressure topping the list. Gallup noted that the intensity of opposition "means that proposed data centers are likely to spur grassroots activism from local residents as well as legal challenges."

Government-mediated wage negotiations between Samsung Electronics and the National Samsung Electronics Union (NSEU) – Samsung's largest union, representing over 45,000 workers – collapsed after two days of marathon talks. The NSEU has authorized an 18-day general strike beginning May 21, 2026, at Samsung's Pyeongtaek semiconductor campus in South Korea. The dispute centers on bonus structures tied to AI-driven semiconductor profits: the union is demanding that Samsung scrap its existing bonus cap, allocate 15% of operating profit to worker bonuses, and formalize those terms in contracts. Samsung's shares fell as much as 6.1% in Seoul following the breakdown in talks. Analysts estimate an 18-day action could cost Samsung between $6.9 billion and $11.7 billion in direct losses, with larger indirect costs. A single one-day action in April produced a 58% production drop for one shift, providing a reference point for the potential scale of disruption.

The bonus dispute reflects a structural divide within Samsung's operations. Memory chip operations – particularly production of HBM, or High-Bandwidth Memory, a specialized chip type required by AI servers – have driven strong profits as DRAM contract prices surged 90–95% quarter-over-quarter in Q1 2026, with TrendForce forecasting a further 58–63% increase in Q2. Meanwhile, Samsung's logic and foundry operations have been loss-making. Workers in the profitable memory division cite SK Hynix's reported bonus of ~$900,000 per employee compared with Samsung's $340,000 offer as their benchmark. Samsung's entire 2026 DRAM production is pre-contracted and sold out, meaning any production shortfall translates directly into delayed shipments to hyperscalers and server manufacturers rather than spot-market shortfall. Samsung has filed an injunction with the Suwon District Court seeking to restrict strike activities that could damage equipment or in-process materials; a ruling is expected before May 21. A smaller union representing workers in Samsung's smartphone, TV, and home appliances divisions has reportedly pulled out of the joint strike, reducing its breadth but not its semiconductor-specific impact.