Nadella Testifies On OpenAI Doubts.

Mazen Al Sabbagh

— 7 min read
Share

TL;DR

  • Nadella Testifies in Musk v. Altman Trial: Microsoft CEO Satya Nadella appeared in court Monday, addressing 2018 emails that reveal significant internal skepticism regarding OpenAI's early trajectory and its relationship with Microsoft. (NewsBytes, AFP/Free Malaysia Today)
  • OpenAI Launches $4B Deployment Entity: OpenAI established its "OpenAI Deployment Company" with a $4 billion initial investment and a $14 billion valuation, signaling a strategic shift into enterprise integration through the acquisition of Edinburgh-based Tomoro AI. (Reuters, Axios)
  • Google Confirms First AI-Built Zero-Day Exploit: Google disclosed the first documented instance of AI-generated zero-day exploits, marking a critical escalation in cyber warfare by criminal hackers. (NYT, Bloomberg)
  • Samsung Strike Looms as Talks Fail: Samsung and its primary union failed to reach an agreement in Sunday talks, increasing the probability of a May 21 general strike with significant implications for global memory supply chains. (DigiTimes, Investing.com)
  • Trump-Xi Summit to Address AI Chip Controls: The upcoming Trump-Xi summit in Beijing will formally address AI chip export controls, bringing semiconductor industry leaders to the negotiating table amid escalating tech competition. (NPR, Bloomberg)

Lead Story: Nadella Testifies On OpenAI Doubts

The most decisive phase of the Musk v. Altman trial commenced Monday in Oakland, with Microsoft CEO Satya Nadella as the opening witness. His testimony centered on a tranche of 2018 emails exposing Microsoft's internal perspectives on OpenAI prior to its initial $1 billion investment, and the subsequent role of that capital in OpenAI's structural transition from a nonprofit to a for-profit entity.

These emails present a nuanced narrative, serving different arguments for each side. Microsoft CTO Kevin Scott, for instance, expressed "highly skeptical" views on an imminent AGI breakthrough and critiqued OpenAI's treatment of Microsoft as "a bucket of undifferentiated GPUs." Concerns from the PR team about associating the firm with claims of "machines beating humans" were also noted. Yet, Scott concurrently warned against OpenAI potentially aligning with Amazon and discrediting Azure, leading to Microsoft's substantial investment the following year.

Musk's legal team will contend that Microsoft's funding corrupted OpenAI's original nonprofit mandate. OpenAI’s lawyers will counter that the investment was purely commercial, and that Musk himself advocated for a for-profit model before departing due to a failure to secure control.

The trial schedule has expedited, with witness testimony anticipated to conclude by May 13 and closing arguments scheduled for May 14. Sam Altman is now slated to testify Tuesday or Wednesday, marking his first sworn public statement since the trial began. Ilya Sutskever, the former chief scientist who voted to oust Altman in November 2023, is also expected to appear before the defense rests. The New York Times captured the courtroom's dynamic Monday, noting the use of props, discernible tension, and the surreal spectacle of two figures with a combined net worth exceeding $670 billion disputing perceived betrayals. (NewsBytes, AFP, NYT, Business Insider, MIT Technology Review)

In Other News

OpenAI Launches $4 Billion Deployment Company, Acquires Tomoro. OpenAI announced Monday the formation of the OpenAI Deployment Company, a majority-owned entity capitalized with $4 billion in initial investment and valued at $14 billion. TPG leads a consortium of 19 investors, including consulting firms and system integrators. This move includes the acquisition of Edinburgh-based Tomoro AI and its 150 forward-deployed engineers, establishing an immediate operational capacity for embedding AI systems within enterprises. The Palantir-style deployment model, which sees engineers integrated into client organizations to customize AI solutions using OpenAI's stack, mirrors Anthropic's recent launch of a competing services arm. This strategic maneuver confirms that leading model providers are aggressively moving up the value chain, directly contesting the enterprise AI services market traditionally dominated by Accenture and the Big Four. (Reuters, Bloomberg, Axios, Yahoo Finance)

Google Reports First AI-Generated Zero-Day Exploit. Google's Threat Intelligence Group published findings Monday confirming the first documented instance of criminal hackers leveraging artificial intelligence to independently discover and weaponize a previously unknown software vulnerability. This zero-day exploit, targeting a widely-used open-source web administration tool, was narrowly disrupted, averting a potential mass exploitation event. While specific details on the cybercrime group and target were withheld, researchers emphasized that AI was instrumental in identifying the flaw itself, moving beyond mere attack refinement. The Register noted GTIG's characterization of this as proof that AI-powered hacking has advanced significantly beyond rudimentary tactics. This development contextualizes recent discussions, including Dario Amodei's warnings regarding Anthropic's Mythos model and the EU Commission's confirmed dialogue with leading AI labs on advanced cyber models, with Anthropic reportedly withholding Mythos from European regulators. (NYT, Bloomberg, Politico, The Register, Reuters/EU Commission)

Samsung Strike Talks Stall as Memory Rivals Rally. Negotiations between Samsung Electronics and the National Samsung Electronics Union concluded Sunday without a resolution, making an 18-day general strike commencing May 21 increasingly probable. Mediated by South Korea's National Labor Relations Commission, the talks failed to bridge the structural divide over bonus allocation. The union seeks 15% of operating profit as uncapped bonuses, aligning with SK Hynix's compensation in peak years, while Samsung management offers 10% with a three-year codification and a one-time $340,000 bonus, which the union rejected. With 30,000 to 40,000 of 73,000 members expected to strike, JPMorgan estimates a potential loss of over 40 trillion won from annual operating profit. Seoul Economic Daily published an analysis suggesting the strike could inadvertently benefit SK Hynix in the DRAM and HBM markets by tightening supply, thus reducing the very profit pool it seeks to expand. Markets are already pricing in disruption: Micron surged 6% Monday, and SK Hynix rallied, as investors anticipate constrained memory supply—a critical factor given HBM chips are fully sold out through 2026. (DigiTimes, Korea Herald, Seoul Economic Daily, Investing.com, Tom's Hardware)

Trump-Xi Summit Will Put AI Chip Controls on the Table. When President Trump arrives in Beijing on May 14 for the first US presidential visit to China since 2017, AI will be formally on the diplomatic agenda alongside established discussions on trade, Taiwan, and Iran. The U.S. delegation includes prominent industry leaders such as Nvidia CEO Jensen Huang, Apple CEO Tim Cook, and executives from Qualcomm, Blackstone, Boeing, and Visa, reflecting the direct economic impact of current export controls. Huang has openly detailed the damage to Nvidia's business, with China revenue "effectively dropped to zero" from $25 billion (32% of total revenue) in 2024. Concurrently, Huawei projects $12 billion in AI chip revenue this year, a 60% increase from 2025. A Sunday analysis by the CFR posits that China holds strategic leverage, controlling rare earth processing and battery component supply chains, and demonstrating a willingness to employ export controls as a retaliatory measure. The summit's outcome will dictate the trajectory of chip trade policy for the remainder of the year, determining whether it yields a framework for bilateral AI safety dialogue or merely defers intractable disagreements. (NPR, Bloomberg, CFR)

X / Social Pulse

The OpenAI Deployment Company announcement generated significant discourse Monday, with debates centering on its validation of, or threat to, the existing AI services landscape. Comparisons to Palantir's forward-deployed engineering model were frequent, though many noted OpenAI’s stronger product foundation and broader ecosystem. Anthropic's concurrent launch of a similar services arm further reinforced the perception of model providers aggressively moving up the value chain. Google’s zero-day disclosure resonated deeply within the cybersecurity community, validating long-standing concerns about AI-powered vulnerability discovery and directly linking to Amodei’s recent Mythos warnings. This added regulatory discussions with the EU Commission's confirmed engagement on advanced cyber models. The Musk-Altman trial continued to attract attention, particularly the NYT's vivid courtroom reporting and Business Insider's outline of remaining critical testimonies, including Sutskever's.

One to Watch

Anthropic's $50B Round: Clock Is Ticking. Anthropic's anticipated $50 billion funding round, poised to value the company at over $900 billion and position it as the most valuable private entity globally, faces an imminent deadline for closure by mid-May. While the company's annualized revenue approaching $40 billion underpins this valuation, and secondary market trading has indicated a $1 trillion implied valuation earlier this month, the official close remains unconfirmed. This situation echoes past frictions in massive private AI rounds, such as SoftBank's experience with OpenAI's margin loan, where private valuations met lender scrutiny. The outcome this week will clarify the market's appetite for such unprecedented private valuations ahead of a potential October IPO. (TechCrunch, The AI Insider)

Quick Hits

  • AMD handed a 2nm CPU order to Samsung over TSMC, signaling a potential diversification in US chipmaker supply chains and a shift in foundry dominance for AI-era chips, according to DigiTimes. (DigiTimes, WCCFTech)
  • Alphabet filed for its first yen-denominated bond sale, joining other tech giants in leveraging international debt markets to finance significant AI infrastructure investments, following a raised 2026 capex guidance of $180-190 billion. (Reuters, Nikkei Asia)
  • Seventy-six percent of large organizations now employ a Chief AI Officer, a substantial increase from 26% in 2025, though cultural resistance remains the primary adoption barrier for 93% of these entities. (CNBC)
  • State AI legislation advanced in multiple jurisdictions: Colorado's AI Act replacement (SB 189) is nearing passage, New York's Assembly passed an AI Training Data Transparency Act, and Vermont moved forward with neurological rights legislation. (Troutman Pepper Locke)
  • Google I/O on May 19 is expected to unveil Gemini 4 and the "Remy" personal AI agent, described in leaked documents as an autonomous "24/7 digital partner" capable of making purchases and sending messages. (Android Authority)

This week is front-loaded with catalysts poised to reshape market dynamics and redefine strategic narratives. Nadella's testimony Monday initiated the trial's critical phase, preceding Altman and Sutskever's appearances. OpenAI’s new deployment company, paralleling Anthropic’s recent services launch, underscores a collective pivot by leading labs towards deeper enterprise integration, moving beyond mere API provision. Google’s disclosure of an AI-generated zero-day exploit concretizes the cyber threat, validating prior warnings. The Samsung labor dispute approaches its flashpoint, threatening global chip supply. Concurrently, the Trump-Xi summit will directly engage AI chip policy. Underpinning these events, Anthropic's pending $50 billion funding round could, if closed this week, establish an unparalleled private valuation, setting the stage for a landmark IPO.

Sources

Lock in. M. mazen@thorterminal.com

Read more