CxO · Monthly · March 27, 2026

the defense-aligned ai fault line

Palantir’s Maven AI is being formalized as a core U.S. military system, with Pentagon funding jumping to around $13B this year and locking it in as long-term infrastructure.[1]

Systems from Palantir and Anthropic’s Claude were used together to identify and prioritize over 1,000 targets in the first 24 hours of strikes on Iran.[2]

At the same time, the Pentagon has labeled Anthropic a 'supply‑chain risk,' threatened to use the Defense Production Act to strip its safety features, and triggered bans on Claude across U.S. federal agencies.[3]

Anthropic is suing the Pentagon over that designation, backed by nearly 150 retired judges and amicus briefs from workers at OpenAI and Google who frame the blacklist as retaliation for refusing unrestricted military use.[4]

OpenAI, xAI, and Google are moving the other way, signing agreements to run their models on classified networks for the Department of War and Pentagon agents, even as a 'Cancel ChatGPT' campaign and a 295% spike in ChatGPT uninstalls signal visible consumer backlash.[5]

the data-center and energy arms race

SoftBank is planning what it touts as the world’s largest data center, a $500B AI complex in Ohio.[6] The site is designed for 10 gigawatts of capacity and will be paired with a $33B natural‑gas plant roughly equivalent to nine nuclear reactors.[7]

Amazon is projecting AI‑driven capital expenditures of $770B in 2026, while Oracle plans to spend tens of billions on AI data centers and GPUs despite its heavy debt load.[8]

Senator Bernie Sanders has responded with a bill to pause new AI data centers in the U.S. and to ban exports of compute hardware to countries without comparable safeguards, explicitly casting AI infrastructure as an existential risk.[9]

The White House has extracted voluntary commitments from data‑center operators to pay for the power they consume, but enforcement is weak and communities in places like Ohio are already pushing back against large facilities.[10]

Meanwhile, Iranian drone strikes that damaged three AWS facilities in the UAE and Bahrain show that physical attacks on cloud infrastructure are now an operational, not hypothetical, risk.[11]

compute supply chains under legal fire

U.S. export controls and GPU demand are now colliding in courtrooms and criminal dockets. Supermicro co‑founder Yih‑Shyan 'Wally' Liaw was arrested for allegedly smuggling $2.5B of Nvidia GPUs to China, following earlier export‑control allegations tied to Iran, and the stock sold off sharply on the news.[12]

Separately, three men have been charged with conspiring to smuggle U.S. artificial‑intelligence technology to China, underscoring pressure on restricted buyers to bypass controls.[13]

On the policy side, the Pentagon’s decision to brand Anthropic a supply‑chain risk and threaten Defense Production Act powers over its models shows that access to secure compute can now be conditioned on political alignment as well as technical security.[14]

All of this sits on top of structural chokepoints like TSMC, which fabricates 90% of the world’s most advanced logic chips including every Nvidia GPU, and fresh warnings that a shutdown of Qatar’s helium exports could disrupt the chip supply chain in the very near term.[15]

identity law as infrastructure moat

California’s Digital Age Assurance Act now forces every operating system, including Linux and SteamOS, to collect a user’s age at account setup and share that information with apps that request it.[16]

Similar age‑verification mandates are rolling out or proposed in more than 25 U.S. states and in Brazil, built from model template bills that shift child‑safety responsibility from apps to OS vendors and device makers.[17]

Meta has spent over $2B lobbying for these laws, which effectively make robust identity verification a regulatory requirement and a potential revenue stream for the few firms able to implement it at scale.[18]

The response is bifurcating: mainstream platforms like Ubuntu plan to comply while 'privacy‑first' projects like GrapheneOS and MidnightBSD openly refuse, even updating licenses to block users in age‑verification jurisdictions such as California and Brazil.[19]

Regulators are not blind to the tradeoffs—FTC officials have acknowledged that age verification can itself violate children’s privacy laws—but have so far declined to intervene.[20]

ai-branded layoffs and the job ladder

Big employers are explicitly tying mass layoffs to an 'AI pivot' even as some engineering demand recovers. Oracle plans to cut up to 30,000 jobs while pouring tens of billions into AI data centers and GPUs.[21]

The company is already carrying more than $100B in debt, amplifying the financial risk of that bet.[22] Meta is weighing layoffs that could hit about 20% of staff, roughly 16,000 people, as AI infrastructure costs mount, while Atlassian has already cut 1,600 roles, about 10% of its workforce, to 'self‑fund' an AI‑first strategy.[23]

Jack Dorsey’s Block laid off around 4,000 employees—nearly half its headcount—explicitly to 'move faster with smaller teams using AI.'[24] Amazon cites AI as the rationale for 16,000 job cuts in 2026.[25]

An AI Jobs Risk Index estimates 9.3M U.S. jobs could be displaced within the next two to five years, and Anthropic’s CEO projects that half of entry‑level white‑collar roles may vanish in three years, yet software development postings are up 15% since mid‑2025 and AI roles are growing.[26]

Workers report that tools like Amazon’s internal AI are often increasing their workload rather than automating it away, feeding a narrative that 'AI' is covering for over‑hiring and cost cutting more than genuine productivity gains.[27]