Apollo, Blackstone back Anthropic's $35 billion capacity expansion in new Broadcom tie-up ↗
Anthropic’s compute appetite just got a monster-sized financing push. Apollo and Blackstone are backing a $35 billion expansion built around Broadcom custom chips and networking gear.
The first chunk adds one gigawatt of AI capacity for the Claude Code maker, with Fluidstack running the data-center sites. This is not “more servers” - this is industrial machinery wearing a chatbot hat. (Reuters)
China prepares $295 billion plan to fund nationwide AI buildout, Bloomberg News reports ↗
China is reportedly lining up a vast national AI infrastructure plan: about $295 billion for data centers over the next five years.
The plan leans hard into interconnected computing hubs and local suppliers, with Huawei named as a likely key player. The subtext is loud: less Nvidia, more homegrown silicon - or so it seems. (Reuters)
US power use to beat record highs in 2026 and 2027 as AI use surges, EIA says ↗
AI’s electricity bill is becoming its own plotline. The EIA expects US power consumption to keep climbing, pushed by data centers, electrification, and, yes, the endless hunger of compute.
Commercial power demand is set to overtake residential demand for the first time on record. Quietly mundane, quietly enormous - like a toaster becoming a city planner. (Reuters)
SpaceX aims to launch orbital AI computing tests by end of next year, sources say ↗
SpaceX is pitching something properly sci-fi: AI compute in orbit. Executives reportedly told investors they want initial demonstration systems up by late next year.
The company says Starship could make space-based data centers commercially viable by lowering transport costs. It has also asked regulators for permission to launch up to 1 million data-center satellites, which is... a lot of sky furniture. (Reuters)
Perplexity plans 2028 IPO regardless of Anthropic or OpenAI listings, CNBC reports ↗
Perplexity says it is sticking with its IPO timeline no matter how the bigger AI listings land. That is confidence, or nerves hidden in a very tidy blazer.
CEO Aravind Srinivas acknowledged there would be ripple effects if the other AI IPOs stumble, but the company’s line is simple: build a high-growth business first, float later. (Reuters)
No tech rule exemption for Apple, EU regulators say amid spat over Siri AI delay ↗
Apple and EU regulators are now openly sparring over Siri AI. Apple blamed EU tech rules for delaying the upgraded assistant in the bloc, citing privacy and security concerns.
The European Commission pushed back hard, saying the rollout decision was Apple’s and that an exemption from interoperability obligations was not on the menu. Basically, both sides are pointing at the rulebook - angrily. (Reuters)
Advancing AI Regulation in Healthcare: Insights from AI Airlock Phase 2 ↗
The UK’s medicines regulator published fresh findings from its AI Airlock work, focused on how AI medical devices should be tested and monitored.
The programme looked at seven technologies, including clinical note-taking, cancer diagnostics, rare eye disease detection, and obesity support. The big takeaway: AI medical oversight needs to follow products through their lifecycle, not just tick a box before launch. (MedRegs)
FAQ
Why are AI companies spending so much on AI infrastructure?
AI companies need vast computing capacity to train, run, and scale advanced models. In this update, Anthropic’s $35 billion expansion shows AI infrastructure moving beyond standard server upgrades and into major industrial buildouts. The combination of custom chips, networking gear, data centers, and financing reflects how compute has become a central bottleneck for AI growth.
What does Anthropic’s Broadcom tie-up mean for Claude Code?
Anthropic’s deal points to a much larger compute base for products such as Claude Code. The expansion is built around Broadcom custom chips and networking equipment, with Fluidstack operating the data-center sites. In practical terms, this suggests Anthropic is preparing for heavier workloads, more users, and more demanding AI development tools.
How could China’s $295 billion AI plan change the global AI race?
China’s reported plan would fund a national AI infrastructure buildout over five years, centred on data centers and connected computing hubs. It also appears designed to depend more heavily on local suppliers, with Huawei named as a likely key player. That could deepen the split between US-led and China-led AI supply chains.
Why is AI increasing electricity demand?
AI systems need large data centers, and those facilities consume substantial amounts of power for computing and cooling. The EIA expects US electricity use to keep rising in 2026 and 2027, supported by data centers and broader electrification. Commercial power demand becoming larger than residential demand would mark a significant shift in how electricity grids are used.
Is space-based AI computing realistic?
SpaceX is reportedly aiming to test orbital AI computing systems by late next year. The idea depends heavily on Starship reducing the cost of launching hardware into orbit. It is still early, but the proposal shows how far companies may go to find new locations for compute capacity, especially as land, power, and cooling become constraints.
What is the UK AI Airlock programme for healthcare?
The UK AI Airlock programme is exploring how AI medical devices should be tested, regulated, and monitored. Its latest phase looked at technologies including clinical note-taking, cancer diagnostics, rare eye disease detection, and obesity support. The key idea is that healthcare AI oversight should continue through a product’s lifecycle, rather than stopping after initial approval.