Anthropic's draft prospectus showed a first-half loss larger than the annual revenue of most listed companies, although most of it was an accounting consequence of the company's own rising value, while one of Blackstone's most senior AI investors argued that the debt piling up beneath the data-centre boom is far sturdier than its critics allow. Americans have placed AI among their ten greatest fears for the first time, and at Google the unease has reached the company's own engineers, some of whom have resigned over its military contracts even as the chief executive describes handing a good deal of his daily reading to a team of AI agents.
1. Anthropic's $80 Billion Loss Is Mostly a Measure of Its Success
Anthropic lost $80.7 billion in the first half of 2026, compared with $13 billion in the same period last year, according to a draft IPO prospectus reported by the Wall Street Journal and reviewed by Reuters. Some $72 billion of that was an accounting charge rather than money spent: the company raised much of its funding through bonds that convert into shares, and as Anthropic's valuation climbed, the estimated value of those bonds rose with it, which the rules require to be booked as a loss.
The cash costs were real enough, with spending on computing capacity reaching nearly $9 billion in six months, against $2.3 billion a year earlier and $7.3 billion for the whole of 2025, and the Journal reported that the company may commit as much as $750 billion more to computing in the coming years. Revenue, however, is growing at a pace that makes those numbers easier to read, having risen to $11.6 billion in the second quarter and to $16.8 billion in the third, a 45 percent quarterly increase that Anthropic disclosed to prospective investors on 14 October.
What this means for you:
The headline loss says more about the mechanics of convertible debt than about Anthropic's trading position. The figures worth watching are revenue growth against compute spending, because a supplier of critical AI tools that cannot cover its infrastructure costs means that prices and terms may change moving forward.
2. Blackstone's AI Chief Says the Debt Fears Are Overblown
Jas Khaira, who runs Blackstone's San Francisco AI unit, BXN1, told TechCrunch Disrupt that worries about the debt financing the AI build-out bear little resemblance to 2008. Blackstone, with $1.3 trillion under management, now describes itself as the largest owner of computing capacity outside the hyperscalers, and it helped invent the model that most lenders now follow: in 2023 it lent CoreWeave $2.3 billion, at a time when the company had about $30 million in revenue, against Nvidia GPUs pledged as collateral and contracts with customers such as Microsoft.
CoreWeave's debt has since grown to about $50 billion and its shares have fallen more than 20 percent in recent days, while the Australian GPU provider Firmus, for which Blackstone co-led a $10 billion credit facility, abandoned a planned listing at a valuation of roughly $30 billion. Khaira acknowledged that rising interest rates and wider credit spreads have increased borrowing costs even for Meta and Microsoft, but argued that because most data centres are built for hyperscalers, the debt is ultimately backed by some of the most creditworthy companies in the world. Asked whether Blackstone's own portfolio companies were seeing returns from AI, he said there was "definitely ROI, but it is uneven."
What this means for you:
The largest lenders to the AI build-out are tying their risk to the credit quality of a handful of customers. Smaller firms buying cloud capacity from neoclouds should check who those providers depend on, since the weaker operators are the ones most exposed if borrowing costs keep rising.
3. AI Enters America's Top Ten Fears for the First Time
Chapman University's twelfth annual Survey of American Fears, which included questions about AI for the first time this year, found that 51.9 percent of more than 1,200 respondents fear AI harming the political process, roughly level with the fear of a third world war and with the 51.8 percent who fear a loved one becoming seriously ill. A further 49.7 percent fear AI damaging the economy, which gives the technology two places in the top ten in its first year on the list. Corrupt government officials remained the leading fear for the eleventh year running, cited by 61.1 percent, and with a margin of error of 3.79 points the order of the remaining places is close to a statistical tie.
What this means for you:
Public anxiety about AI is now on the same scale as anxiety about war and illness, and that shapes how customers and employees respond to any visible use of it. Businesses deploying AI in customer-facing work should expect scrutiny and explain plainly what the tools do.
4. Google DeepMind Staff Resign Over Military Contracts
Several engineers have resigned from Google DeepMind, or are in the process of leaving, over the company's work with the US and Israeli militaries, the BBC reported. Sofia Liguori, a London-based research engineer who left this week after three years, spoke to the BBC despite contract terms restricting her from doing so, saying that the company's growing push for AI in military operations made her "ashamed to work at Google" and had cost her faith that it would honour its own stated values.
Those values, set out in Google's AI Principles, commit the company to human oversight and to alignment with international law and human rights, and the departing staff, along with experts the BBC consulted, questioned how much human oversight applies at critical moments. Google is one of several technology companies that signed contracts with the US military earlier this year to deploy AI in classified environments, part of a push by Defence Secretary Pete Hegseth to turn the armed forces into what he calls an "AI-first warfighting force". Google said it follows a consistent review process for public-sector contracts, including its acceptable use policies and AI principles, and that employees have dedicated channels to raise concerns, protected by strict policies against retaliation.
What this means for you:
The question of where a company's AI is used, and under what oversight, has become a reason for skilled staff to leave. Any business that publishes principles about its use of AI should expect employees to hold it to them.
5. Sundar Pichai Runs Google With a Team of AI Agents
Sundar Pichai said on Alex Heath's Sources podcast that he relies on a group of AI agents to keep up with a company of Google's size. The agents summarise the previous 24 hours of work on Google's models, track how computing capacity is being used across product areas through direct access to internal telemetry, and reconstruct discussions that run in parallel across dozens of internal chat groups, so that Pichai no longer needs to read every thread himself. Because they run continuously in the cloud and keep context between sessions, he described their main value as continuity. Google plans to bring the same kind of persistent agent to consumers over the coming months, built into Search and Gmail and able to carry on working after the user has stopped typing.
What this means for you:
The most useful agents at the top of a large company are not making decisions; they are compressing information that no single executive could read in full. Leaders drowning in reports, channels and threads have a practical model to copy, starting with a daily summary drawn from the systems they already use.
About Us
At AI-First Mindset, we help leaders bridge the gap between knowing AI and using it. Our work includes hands-on workshops and operating frameworks designed for companies that want to embed AI into workflows without relying on large institutional vendors.
To explore how we can help, contact us at aifirstmindset.ai.