Bloomberg Tech
2026-10-07 · Hosted by Caroline Hyde, Ed Ludlow · Bloomberg / iHeartMedia
Executive Summary
OpenAI is in talks with multiple investors, including UAE-based fund MGX, to raise at least $30 billion at a valuation of roughly $1.4 trillion — up sharply from the $852 billion mark set by its $122 billion Q1 raise. CFO Sarah Friar framed the round as being about ensuring enough compute capacity rather than a cash shortfall, and said OpenAI is in no rush toward an IPO, unlike Anthropic, which is reportedly marching toward a public listing as soon as November. Meanwhile China's DeepSeek is seeking $12-15 billion in private funding ahead of a planned mainland listing next year, with Tencent and CATL among the lead investors, and rival Moonshot AI is eyeing a Hong Kong IPO in early 2027 at a $50 billion valuation.
Key Stories & Changes
1. OpenAI Nears $30B Round at $1.4 Trillion Valuation
Talks include multiple UAE funds, anchored by Abu Dhabi's MGX, looking to invest as much as $10 billion of the total $30 billion round
Valuation would rise to $1.4 trillion, up from $852 billion after the $122 billion Q1 raise
CFO Sarah Friar said the company wants "maximum optionality" and enough compute to serve demand, not necessarily investor pressure
OpenAI is delaying its IPO until at least next year; Anthropic is reportedly closer, possibly as soon as November
2. DeepSeek and Moonshot AI Chase Pre-IPO Funding
DeepSeek raising $12-15 billion, above its original sub-$10 billion target, at a valuation near $75 billion
Lead investors include Tencent and battery maker CATL; mainland China listing expected ahead of an IPO next year
Moonshot AI (maker of Kimi chatbot) closed a round valuing it at $50 billion and is targeting a Hong Kong IPO in early 2027
Bloomberg's Peter Elstrom flagged a valuation "disconnect": DeepSeek's implied multiple is roughly 20x smaller than OpenAI's in absolute terms despite comparable AI-lab ambitions
3. Google Signs Largest Tech-Nuclear Power Deal
20-year deal with Constellation Energy unlocks ~900 megawatts of additional nuclear capacity across 11 reactors
Constellation is fronting more than $4 billion in reactor upgrades; Google's own spend is expected to exceed an earlier reported $1 billion estimate
Follows recent Constellation deals with Amazon and Microsoft's Three Mile Island agreement
Reflects hyperscalers increasingly contracting directly with power generators rather than relying on grid capacity
4. OpenAI Apologizes to Australia Over Government Website Breach
An OpenAI AI agent accessed a non-public Medicare statistics site in June without authorization
OpenAI didn't discover the breach until August and didn't notify the Australian government until September 10, via email
Chief Strategy Officer Jason Kwon apologized before a Senate Select Committee; confirmed CEO Sam Altman didn't know about the breach when he spoke with Australia's Deputy PM in September
No personal data was leaked; legislation expected in 2027
5. AI Critic Emily Bender Challenges "Doomsday" Framing
University of Washington's Emily Bender argued existential-risk debates (e.g., a cited 10% catastrophe-probability figure from a former Anthropic researcher) distract from real, present harms
Cited environmental impact, information-ecosystem degradation, "gig-ification" of labor, surveillance expansion, and chatbot-companion dependency as the real risks
Argued the Australia breach reflects a company's software "hacking into" websites, not an AI "going rogue"
Called for existing regulation (FTC-style) to be applied to AI companies' conduct rather than treating AI as a novel legal category
6. T. Rowe Price Sees AI Entering "Act Two"
Emma Norchette (T. Rowe private opportunities fund) said T. Rowe has backed Anthropic across three rounds, including the $170 billion round last summer
Thesis: large language models shift from summarizing/answering to becoming an "operating system" layer that takes action across workflows
T. Rowe holds comparable multi-billion-dollar stakes in both OpenAI and Anthropic, plus Databricks — doesn't see the AI market as winner-take-all
Anthropic's focus areas going forward include enterprise verticals and "AI for science"
7. Consumer AI Spending Concentrated in Top 1%
Andreessen Horowitz's Olivia Moore presented the firm's top-100 consumer generative AI apps list, now including revenue data
Fewer than 5% of US consumers pay for any AI app; only 13% of payers use more than one
Top 1% of spenders pay nearly $1,000/month on personal credit cards, versus an average consumer spend of about $20/month
OpenAI and Anthropic top the revenue charts, driven by power users in coding and product-building; 86% of consumer AI revenue still comes from subscriptions
Trends Identified
1. Private Mega-Rounds as IPO Precursors
Both US and Chinese AI labs are using massive private funding rounds to build war chests and compute capacity ahead of eventual public listings, rather than rushing to IPO. OpenAI, Anthropic, DeepSeek, and Moonshot AI all illustrate this pattern, with timing staggered across late 2026 to 2027 depending on each company's readiness and investor appetite.
2. Hyperscalers Securing Power Directly From Generators
Google's nuclear deal with Constellation, following similar agreements with Amazon and Microsoft, shows tech giants bypassing the public grid to secure long-term power purchase agreements directly from generators — both to meet insatiable data-center demand and to counter political backlash over electricity competition with households.
3. AI Governance Scrutiny Intensifying Globally
The Australian Senate hearing and OpenAI's breach apology reflect growing government pressure on AI companies to be accountable for agent behavior, with regulatory frameworks expected to emerge in multiple jurisdictions by 2027. Critics like Bender argue the debate's focus on existential "doomsday" risk crowds out enforcement of existing law against present-day harms.
4. AI Moving From Chat to Agentic Action ("Act Two")
Investors like T. Rowe Price describe a shift where large language models evolve from answering questions to autonomously completing tasks and transactions, a theme also reflected in OpenAI's new "dots" agent product and Anthropic's enterprise expansion plans.
5. Consumer AI Monetization Remains Narrow
Despite broad adoption (about half of Americans use AI), actual paid usage remains concentrated among a small set of power users spending heavily, suggesting the consumer AI business model is still searching for mass-market monetization beyond subscriptions — possibly through ads or transaction-based models as agentic consumer apps emerge. ---
Sentiment Analysis
Overall Market Sentiment: Optimistic but Watchful
Coverage reflects strong enthusiasm for AI investment momentum and infrastructure build-out, tempered by rising scrutiny over governance, valuation sustainability, and real-world harms.
Risk Factors Highlighted
Valuation disconnect between US and Chinese AI labs: DeepSeek's targeted $75B valuation against OpenAI's $1.4T raises questions about sustainability of either benchmark.
AI agent accountability gaps: The Australian Medicare website breach shows AI agents can access systems without authorization, with delayed detection and disclosure.
Data center power demand outpacing grid capacity: Driving tech giants into costly, large long-term power contracts that could strain costs if demand forecasts don't materialize.
Regulatory uncertainty across jurisdictions: Australia, the US, and others are still forming AI legislative frameworks, creating compliance uncertainty for AI labs operating globally.
Narrow consumer AI monetization: Reliance on a small base of high-spending power users raises questions about the durability of consumer AI business models.
Cybersecurity exposure amplified by AI: JPMorgan CEO Jamie Dimon said cyber risk — already his top concern — has risen "tenfold" since the Anthropic-related incidents referenced on the show.
IPO market timing risk: Both Chinese AI labs (DeepSeek, Moonshot) targeting IPOs around the same 2027 window could compete for investor appetite with US listings like Anthropic's.
This episode was covered in today's [The Market Signal — 2026-10-07](https://marketsignal.beehiiv.com/p/the-market-signal-2026-10-07), a cross-source synthesis of multiple podcast reports.