Selection window: absolute time (UTC). Current time 2026-07-14 22:37 UTC, cutoff = 2026-07-13 22:37 UTC. Only stories verified as published after the cutoff (within the last 24 hours) are included below.
US Commerce official: “very few” Nvidia H200 chips have actually shipped to China
Jeffery Kessler, Under Secretary of Commerce for Industry and Security, told a congressional hearing that only “a very small quantity” of Nvidia’s H200 AI chips have been shipped to China and Hong Kong. The Commerce Department cleared roughly 10 Chinese firms — including Tencent and ByteDance — to buy the H200 back in May, so the remark confirms how wide the gap remains between what was approved and what actually crossed the border. That said, it also signals that H200 shipments have restarted at all after a long freeze, which leaves room for competing readings of Nvidia’s China revenue outlook. Initial planning had targeted between 5,000 and 82,000 units before mid-February 2026, a timeline that came and went without a single unit delivered.
reference: CNBC · Bloomberg · South China Morning Post
Dimon: AI has cut some JPMorgan teams by 40% — but the bank won’t run dramatically cheaper
JPMorgan CEO Jamie Dimon said on the bank’s second-quarter earnings call that AI has reduced headcount by as much as 40% in certain areas. He was careful to add that this will not dramatically lower operating costs or lift profit margins — in a competitive market, he argued, every bank ends up spending its AI advantage on better customer service rather than fatter margins. Most affected employees were offered roles elsewhere in the firm. JPMorgan runs close to 1,000 AI use cases across fraud protection, marketing and note-taking on a technology budget of nearly $20 billion, and CFO Jeremy Barnum noted that while AI token expenses are trivial today, they expect meaningful acceleration in the second half. The bank posted net income of $21.2 billion for the quarter, up 41% year over year.
reference: CNBC
Apptio’s co-founders reunite for Thira, a back-office automation startup, with $21M seed
Apptio co-founders Sunny Gupta and Kurt Shintaffer have launched Thira, a Bellevue, Washington-based enterprise AI startup, with $21 million in seed funding led by Madrona and joined by FUSE. Thira’s pitch is a “back-office that runs itself” — AI handling the behind-the-scenes tasks that keep large companies running, like setting up a new hire’s laptop, resetting a locked account, or approving a software purchase. It starts with IT support, building agents that take a ticket, work it across the systems where fixes actually happen — ServiceNow, Jira Service Management, Freshservice, plus the identity and device-management tools connecting them — and close it out; finance and HR are on the roadmap. The team includes Atlassian’s former SVP of engineering along with alumni of Oracle Cloud Infrastructure and Databricks, and it enters a crowded field where ServiceNow (which bought Moveworks) and Aisera already compete, betting on the CIO relationships the founders built at Apptio. Thira is working with 10 design partners ahead of a broader launch this fall.
reference: GeekWire
Today’s takeaway
US chip controls on China remain stuck in an awkward middle state — the approvals exist, but almost nothing has shipped — which keeps the timing of any Nvidia China revenue recovery genuinely unclear.
In finance, a CEO put a hard number on AI displacing work (up to 40% in some JPMorgan units), and then immediately punctured the assumption that this translates into cost savings.
And on the investment side, Thira is one more data point that money is chasing agents that finish boring back-office work rather than frontier models.
Daily AI Brief 
Leave a comment