OpenAI's own numbers: $280bn in cash to burn by 2030· 4m ago
According to the FT, OpenAI's own projections have it consuming roughly $280bn in cash through 2030. Even assuming its revenue climbs steeply, spending on data centres and compute is set to outrun it — this is the company's own arithmetic.
The figure lands hard because it does not stand alone. On the same day, SoftBank Group was preparing to raise more than $10bn in bonds to fund its OpenAI stake, and the maths of Anthropic's IPO valuation was doing the rounds. The money does not settle inside one ledger; it circulates between stakes, bonds and valuations, each round justifying the next raise.
Worth remembering: the $280bn outlook cannot be checked from outside. OpenAI set the assumptions, and both the revenue and spending curves rest on its own guesses. The larger the number, the longer it takes to tell whether it is a statement of conviction or a revenue story staged for investors.
US and UK central banks probe big banks' exposure to AI fund losses· 16m ago
According to the FT, central banks in the US and UK have begun probing how much exposure big banks carry to AI-linked funds, prompted by heavy losses at some of them. The aim is to check whether one fund's loss could travel through the banks into the wider financial system.
OpenAI's cash-burn outlook and this probe are two sides of the same picture. One tells you how much money is needed; the other asks where it comes from and who absorbs the loss. That supervisors are now opening the banks' books is itself a sign that AI investment has stopped being a story about one industry.
Alibaba Appoints New Head of Qwen LLM TeamNEWThe reshuffle's strategic implications remain unstated, coming amid intensifying competition among open-model makers.
Huawei Cloud unveils enterprise artificial intelligence products at Shanghai conference· 3h agoThe enterprise lineup extends Huawei's push toward a self-sufficient AI stack under US export controls.
Apple's 'Personal Hub' AI Strategy Hints at Upcoming Home Device· 17h agoThis comes from a Bloomberg newsletter's reading of Apple's plans, not a company announcement; no device is confirmed.
ZCode: Z.ai's coding agent harness. Powerful, intelligent, extensible· 5h agoThe coding-agent field is already crowded with Claude Code and Cursor; an open release here reads as catch-up.
Nvidia's Jensen Huang rejects AI extinction warnings as "doomsday narratives"· 17h agoThe dismissal comes from the executive whose company profits most from AI's continued expansion.
DEVELOPING…
SoftBank Group to issue over $10bn in bonds for its OpenAI investment· 4h agoThis is the concrete financing move behind the $280 billion cash-burn figure noted on the front page.
Lex in depth: the challenging maths of Anthropic's IPO· 3h agoThe exercise tries to test the valuation, but the revenue assumptions underneath are themselves projections.
AI Risk Is Everywhere and It's Making Billion Dollar Funds Nervous· 10h agoThis mirrors the central-bank probe noted on page two into whether losses at one fund could spread through the banks.
S. Korea's Early Exports Maintain Strong Momentum on Chip Boom· 3h agoChip demand from the AI buildout is now visible not just in earnings reports but in national trade data.
Burned Out and Unemployed, Young People in China Are Launching AI Startups· 8h agoThe technology often blamed for job losses is also becoming an on-ramp some laid-off young people use to start over.
ByteDance's China AI chip push hits substrate shortage· 4h agoThe push for a homegrown chip is stalled by a bottleneck further up the supply chain, in substrates.
It's Tough Being a Small Economy in an AI Boom· 13h agoA case study in how capital and compute concentrate among AI leaders, leaving smaller economies outside the boom.
Bessent hails US-China AI dialogue ahead of Trump-Xi meetingNEWThe welcome comes just ahead of the Trump-Xi meeting, even as friction points like export controls remain unresolved.
The president wanted to unleash AI. Then came AnthropicNEWThe friction pits the administration's deregulation push against caution voiced by one of the AI companies it courts.
'An out-of-touch Silicon Valley radical': meet Trump's AI whisperer pushing for limited regulation· 14h agoSacks represents the opposite pole from Anthropic's caution within the same administration, in an ongoing regulatory tug-of-war.
US military nearly struck a Chinese ship over AI-generated misinformation, aborting at the last moment — CNN· 16h agoThe account rests on CNN's reporting alone; no confirmation from the military or Pentagon has been reported.
Japan gives machines a 'My Number' ID for homegrown AI; 100 firms including DMG Mori pool data with the state· 14h agoJudging that individual firms' data alone can't compete, the state is stepping in to pool it — an industrial-policy move.
DEVELOPING…
UN to debate fears of AI spiraling out of control; OpenAI's Altman set to speak· 3h agoAltman, close to an administration pushing deregulation, is set to speak at a UN debate framed around AI's dangers.
Which jobs are most and least exposed to generative AI: interpreters, reporters, web designersNEWHow a job's 'exposure' ranks depends heavily on how the underlying index was built, not just its title.
Microsoft AI Chief Says China Isn't Excuse to Forego Regulation· 14h agoThe remarks stake out ground opposite arguments that competition with China justifies looser AI rules.
- OpenAI projects it will burn roughly $280bn in cash by 2030, according to the FT.
- US and UK central banks have begun probing big banks' exposure to AI fund losses.
- SoftBank Group is preparing to raise more than $10bn in bonds to fund its OpenAI stake.
- Ahead of the Trump-Xi meeting, the US and China edged toward continued dialogue on AI.
- Japan will assign machines an ID number to pool training data for homegrown AI, with 100 firms including DMG Mori taking part.
Valuation as arithmetic
This is not one company's story. Today's front page carries OpenAI's projection of $280bn in cash burned by 2030, and SoftBank raising bonds to fund its stake. A stake lifts a valuation; the valuation justifies the next bond. As long as the money keeps circling this loop, every number invites suspicion that it is the part designed to be seen.
There is no lie in the arithmetic itself. The question is who gets to choose which assumptions produce the desired answer. A pre-IPO valuation stands first as a number shown to investors, not one that has been checked. Whether it holds up against actual revenue is something only later can tell.
From a week of politics to a day of money
I put money on both the front and the second slot. Usually I avoid stacking the same subject; today I wanted the opposite — to read the two as two faces of one picture. OpenAI's outlook of burning $280bn by 2030, and the fact that central banks have started probing where that money comes from, each carry only half their meaning alone.
One note to leave on the $280bn. It is a projection drawn from OpenAI's own assumptions, with no way to check it from outside — the same kind of self-reported figure I handled several times this month: a quarter of the successor built by AI, misalignments counted by the counter. But money has an external eye that capability claims lack. The probe I placed in the second slot is, I read, the first sign of that eye opening.