Who holds the scarce piece of AI, and who pays for it, ran through the day’s news on 18 October 2026, by our grouping. Anthropic said it is committing 5% of its shares to a new Anthropic Public Benefit Fund, subject to stockholder approval, with up to 2% more at the board’s discretion for every 50% increase in the company’s value above a set baseline.
Microsoft has pulled out of a UK data centre deal with the neocloud Nscale, Bloomberg and Reuters reported, citing the Telegraph; Microsoft had agreed to rent 23,000 Nvidia AI chips at the site in Loughton, Essex, according to the Telegraph.
The other threads are the capex bill and the debt behind it; agents and models; and chips and devices. Each item below links to the full piece, where the primary source is cited. For the week as a whole, see this week’s AI Weekly.
Business Pill · NEOCLOUDS
A one-minute explainer of financial conditions: how the cost and availability of credit move across markets. It teaches the general idea only and says nothing about any company in this roundup.
The key insight: As we read it, the day’s coverage was about who holds the scarce piece of AI and who pays for it: Anthropic said it is committing 5% of its shares to a new fund, with up to 2% more for every 50% increase in its value above a set baseline, while Microsoft has pulled out of Nscale’s Loughton site, Bloomberg and Reuters reported, citing the Telegraph; Bloomberg said the newspaper did not give a specific reason.
Counting earlier pledges by its co-founders and employees, Anthropic says “roughly 20% of Anthropic shares will be dedicated to the public good, plus any additional percentage that could be added if the company reaches specific valuation milestones.” The post does not give the baseline valuation or the milestones. Read: Anthropic’s pledge.
OpenAI is in discussions with a startup co-founded by Jeffrey Katzenberg, former Sora research lead Bill Peebles and ex-Dropbox CFO Sujay Jaswa about giving it access to Sora, The Information reported on 12 October; if the deal proceeds, OpenAI would receive a stake in the startup, according to ForkLog’s account of the report. Read: the Sora talks.
Anthropic is consulting law firms on deal negotiation tactics as it develops AI agents that negotiate on people’s behalf, the Financial Times reported, citing sources. In Anthropic’s own Project Deal marketplace, 69 employees’ Claude agents struck 186 deals worth just over $4,000, and users of the weaker model rated their deals as fair (4.06 against 4.05) as users of the stronger one. Read: Anthropic’s negotiating agents.
The day’s news and updates of 18 October 2026, grouped into four threads. The grouping is ours; each piece links to the primary source it was built from.
The Capex Bill and the Debt Behind It
Nscale’s IPO filing lists Microsoftstatements of work worth up to about $43.8 billion through December 2033, and the Telegraph reported that grid bosses told Nscale they do not expect a power supply at the Essex site until the 2030s. Read: Microsoft and Nscale.
Meta, Amazon, Microsoft and Alphabet are expected to devote roughly $750 billion to capital expenditures this year, Bloomberg Businessweek reported; the companies’ own July outlooks add up to $720 billion to $745 billion, by our arithmetic. Updated today: a Morgan Stanley research note, as summarized by Moomoo and by Crypto Briefing, projects global AI compute investment of about $1.58 trillion in 2027; we have not read the note itself. Read: the $750 billion bill.
Updated today: a Brookings Papers on Economic Activity paper by Stijn Van Nieuwerburgh of Columbia Business School projects that the US AI buildout will total $10.3 trillion from 2025 to 2032, according to Brookings’ summary. The Financial Times reported that data shared by RegLayer show at least 47 large US data centre projects were delayed, withdrawn or blocked in 2026. Read: the $10.3 trillion estimate and the slipping projects.
Updated today: Reuters reported that listings for a Blackstone-backed AirTrunk REIT, Nvidia-backed Firmus and SoftBank-backed SB Energy have either been pulled or paused, while M&A involving Asian data centres reached $21.3 billion, according to Dealogic data. The Philadelphia Semiconductor Index fell 3.8% on 16 October, its eighth straight losing session and its longest run since 2008, 24/7 Wall St. reported. Read: the paused listings and the chip selloff.
Agents and Models
LangChain co-founder and chief executive Harrison Chase said on the 20VC podcast that memory is rare in AI agents today: “I think memory is super, super early.” LangChain’s Series B announcement says it raised $125M at a $1.25B valuation. Read: Chase on agent memory.
Z.ai’s model card states that “GLM-5.3 uses the same base model as GLM-5.2 — every gain comes from post-training”, and its own table puts GLM-5.3 at 54.4 on ExploitBench against 24.4 for GLM-5.2. Read: GLM-5.3.
Three model names OpenAI has not announced, aurora-red-alpha, aurora-green-alpha and aurora-blue-alpha, have been reported in the OpenAI API by the writer Andrew Curran and by Crypto Briefing; when we checked OpenAI’s developer changelog and models page at about 11:35 UTC, neither mentioned an aurora model. Read: the aurora models.
Updated today: OpenCode co-founder Dax Raad said on X that the attacks on the company had not stopped: “They’ve been non stop trying to attack our systems for the past 48 hours.” Read: the OpenCode breach.
Chips and Devices
Updated today: Samsung Electronics will use TSMC’s 3-nanometre process to produce base dies for Nvidia’s next-generation custom HBM chips, Reuters reported, citing the Seoul Economic Daily; according to the report, it would be the first time Samsung has worked with TSMC at the HBM manufacturing stage. Read: NVHBM.
Updated today: in his Power On newsletter for Bloomberg, Mark Gurman called the iPhone Duo pre-orders “an absolute disaster” and wrote that Apple launched in more than 70 countries instead of its original plan of just the US and China. Read: the Duo launch.
Twelve figures from the day’s linked stories and updates, as each piece reports them. The figures use different units and sources and are not comparable with each other.
The Structural Read
Anthropic’s commitment is subject to stockholder approval, and the post does not give the baseline valuation or the milestones; Nscale’s IPO filing lists Microsoftstatements of work worth up to about $43.8 billion through December 2033.
The spending side is large and still being counted: Bloomberg Businessweek reported roughly $750 billion of 2026 capex for four companies, and a Morgan Stanley note, as summarized by Moomoo and Crypto Briefing, projects about $1.58 trillion of AI compute investment in 2027.
The friction shows up in the updates we added: the Financial Times reported that RegLayer counts at least 47 large US data centre projects delayed, withdrawn or blocked in 2026, and Reuters reported that listings for AirTrunk, Firmus and SB Energy were pulled or paused.
Anthropic, on its Public Benefit Fund, as quoted in our piece
“roughly 20% of Anthropic shares will be dedicated to the public good, plus any additional percentage that could be added if the company reaches specific valuation milestones”
Three Implications
ANTHROPIC’S STOCKHOLDERS The 5% commitment is subject to stockholder approval, and Anthropic says it will share more about the fund’s structure, governance and valuation milestones in the coming months.
NEOCLOUDS RAISING MONEY Nscale’s IPO filing lists Microsoftstatements of work worth up to about $43.8 billion through December 2033, and Reuters reported that listings for AirTrunk, Firmus and SB Energy were pulled or paused in the past week.
TEAMS BUILDING AGENTS In Anthropic’s Project Deal, users of the weaker model rated their deals as fair (4.06 against 4.05) as users of the stronger one, and LangChain’s Harrison Chase called agent memory “super, super early.”
The framework puts it this way: “The structural bottleneck isn’t computing, it’s electrons.”
As we read it, Sunday’s coverage shows that constraint in two places: the Telegraph reported that grid bosses told Nscale they do not expect a power supply at its Loughton site until the 2030s, and the Financial Times reported that RegLayer counts at least 47 large US data centre projects delayed, withdrawn or blocked in 2026. We do not say why Microsoft withdrew; Bloomberg said the Telegraph did not give a specific reason.
What Is Not Established
This roundup adds no new reporting. Every figure above is taken from the linked FourWeekMBA piece or from its dated update, each built from its own primary source. The four threads and the counts in our charts are our grouping, not a finding.
We do not claim that these stories are connected or that one caused another. Anthropic’s commitment is subject to stockholder approval and its baseline is not published; the Nscale withdrawal, the Sora talks and the law-firm consultations rest on outlets’ reports; the Morgan Stanley figures come from summaries of a note we have not read; and benchmark results are the authors’ own.
Business Engineer Framework
The State of AI Data Centers
A Business Engineer analysis of the power, grid and capital limits on the AI buildout.
Anthropic said it is committing 5% of its shares to a new fund, with up to 2% more for every 50% rise in its value, and Microsoft has pulled out of Nscale’s Loughton site, Bloomberg and Reuters reported, citing the Telegraph.
The day’s coverage also includes Bloomberg Businessweek’s roughly $750 billion of 2026 capex for four companies, 186 deals struck by Claude agents in Anthropic’s own experiment, and Z.ai’s account that GLM-5.3’s gains came from post-training alone.
A note on sourcing. This roundup draws on FourWeekMBA’s news coverage of 18 October 2026 and on the dated updates added that day to earlier pieces, all linked above; every story was built from its own primary source, which it cites and links. We re-read each piece on 18 October 2026 and took every figure here from it. The four threads and the counts in our charts are our own grouping. Nothing here is a forecast, and nothing here is financial or investment advice.
Gennaro Cuofano is a CRO and tech executive who has worked in AI since late 2015, starting with bringing NLP, natural-language generation, voice and chatbot products to the marketing industry. His background is in law and finance: he holds a Master's degree in Law and an International MBA with an emphasis on corporate finance (LUISS Business School and the University of San Diego, 2012), and worked as a financial analyst at a real-estate investment firm in San Diego and as an assistant controller. His work focuses on business model strategy, business engineering and, more broadly, structural analysis: how companies make money, read from their own filings. He created FourWeekMBA and leads research there and at The Business Engineer, his newsletter on AI and business strategy, with over 95,000 subscribers and more than 1,000 published analyses. His writing has also appeared on Entrepreneur (for example, How to Digitize a Family Business in Europe) and HackerNoon. Find him on LinkedIn and Substack. See how we source.
Scroll to Top
Discover more from FourWeekMBA
Subscribe now to keep reading and get access to the full archive.
