CxO · Weekly · May 4, 2026

frontier ai economics and shifting alliances

OpenAI missed internal revenue and user targets, projects a $14B loss in 2026 despite over $20B in revenue, and its CFO is publicly questioning whether it can pay future compute bills if growth stalls.[1]

Microsoft and OpenAI have ended their exclusive revenue‑sharing deal, knocking Microsoft’s stock about 5% and freeing OpenAI to sell into Amazon Bedrock and potentially Google.[2]

ChatGPT’s share of generative‑AI web traffic dropped from 86.7% to 64.5% in a year, while Google’s Gemini climbed from 5.7% to 21.5%.[3]

Meanwhile Anthropic reports $44B in ARR, a valuation near $1T, and is asking investors for roughly $50B more within weeks, while xAI is sitting on 550,000 Nvidia GPUs but reportedly using only 11% of that capacity.[4]

GitHub Copilot and other tools are shifting to metered billing amid a cost crunch, and DeepSeek has slashed API prices by up to 90%, underlining pressure on premium model pricing.[5]

defense, drones, and humanoid robots as ai demand sinks

The Pentagon increased its annual drone budget from $225M to $55B in a single year and signed classified any‑lawful‑use AI deals with Google and other labs, plus separate integration agreements with Microsoft, OpenAI, and Nvidia.[6]

The US Navy added a $100M contract for AI‑trained underwater drones that can update mine‑detection algorithms in days instead of months.[7] Ukraine plans to deploy 25,000 ground robots to replace soldiers in frontline logistics, while China is spending $1B to swap 8,500 human maintenance workers for AI robots.[8]

Companies like Figure AI and 1X are scaling humanoid production from one robot per day to 24 per hour and targeting 10,000 units from a single factory, with SoftBank planning a robotics‑driven data‑center business and a hoped‑for $100B IPO.[9]

At the same time, Palantir is seeing its policing AI called intrusive, rejected by the German military, and investigated by London’s Metropolitan Police, even as it lands a $300M USDA deal.[10]

us–china tech and capital decoupling

Nvidia now reports zero percent legal accelerator market share in China due to US export controls, even as its GPUs are reportedly smuggled in for local AI projects.[11]

China has fully fenced off its AI sector from US capital, instructed its top labs to reject American investment, and blocked Meta’s $2B acquisition of AI startup Manus on security‑leak grounds.[12]

In parallel, Beijing is funding an $8.4B orbital data‑center constellation and a 1GW space‑based computing platform, while pushing domestic champions like Zhipu, which is processing 5.5T tokens per day, and DeepSeek V4, now roughly eight months behind US frontier models.[13]

China has open‑sourced trillion‑parameter model Ling‑2.6‑1T that reportedly burns fewer tokens than US peers and rolled out strong labor rulings making it illegal to fire workers just to replace them with AI.[14]

On the US side, the FCC has banned Chinese labs from certifying electronics—which affects about 75% of US‑bound devices—while California is using loopholes to constrain Chinese EVs even as its own high‑speed rail cost estimate blows out to $231B.[15]

enterprise ai roi, costs, and overbuilt infra

Nvidia’s vice president of applied deep learning says the cost of AI compute is currently higher than employing human workers, and an MIT study finds automation is economically viable in only about 23% of jobs.[16]

Big Tech is still projected to spend roughly $700–805B on AI and data‑center capex while announcing 81,747 tech layoffs in Q1 2026 alone and blaming some of them on AI.[17]

Uber burned its entire 2026 AI coding budget in four months, with 95% of engineers using AI tools and 70% of committed code originating from AI, yet it now faces hard questions about efficiency and ROI.[18]

Oracle and other cloud players are sitting on idle data‑center capacity even as half of planned US data‑center projects are delayed or canceled and a new Utah AI facility is expected to consume more than twice the power used by the entire state.[19]

Environmental pushback is growing as gas‑powered data centers are projected to emit more greenhouse gases than some countries, and US power demand is forecast to hit record highs in 2026–2027, driven largely by AI.[20]

regulators, cars, and the ai backlash

California will start ticketing driverless cars for traffic violations in July just as emergency responders report increasing interference from Waymo vehicles during critical incidents and riders complain about basic service failures like cars driving off with luggage.[21]

GM has pushed Google’s Gemini into four million recent‑model vehicles via OTA, replacing Google Assistant and moving to monetize driver data, while vocal segments of drivers say they preferred CarPlay/Android Auto and are threatening to switch brands.[22]

Utah has become the first US state to hold websites liable when users mask their location with VPNs to bypass age checks, and the EU has formally found Meta in breach of child‑protection law for failing to keep minors off its platforms.[23]

The FCC is reviewing ABC broadcast licenses after a Jimmy Kimmel joke offended political leaders, while simultaneously voting to ban Chinese labs from certifying electronics and approving Starlink rules that could raise its capacity sevenfold.[24]

Cultural institutions are reacting too: the Oscars have banned AI‑generated actors and writers from awards, and Bernie Sanders is hosting a high‑profile event focused on AI existential risk and global controls.[25]