Grantham warns $175B AI bubble risks global collapse

Markets (2)

• Quinn notes a four standard deviation unwind in the momentum factor coinciding with yen intervention overnight.
• Quinn points out that triple-levered ETFs drove $100 billion into the AI semiconductor trade, a movement he considers completely removed from fundamentals.

AI Infrastructure (1)

• Eric Glenn argues the AI trade is cracking, citing Meta's potential sale of excess compute and a viral tweet predicting a memory efficiency breakthrough as catalysts that challenged reflexive positioning.

Macro (1)

• Eric Glenn sees gold as a favored trade to express fading Fed hawkishness, noting real yields are peaking and cyclical inflation components are rolling over.

Labor (1)

• Quinn interprets the mixed jobs report as a peak in re-acceleration, citing a miss in NFP and falling labor force participation despite a lower unemployment rate.

Fed (1)

• Eric Glenn asserts there is no case for Fed hikes, arguing wage growth shows no signs of revitalization and the Fed should not hike into a tepid, supply-constrained labor market.

Models (1)

• Nathaniel Whittemore reports hints from AI leaker M1 Astra suggest Anthropic's Fable model relaunch will require credit-based usage and user identity verification. Max Weinberg and Dario Amodei previously compared advanced AI access to gun licenses.

Regulation (1)

• Senator Mark Warner is drafting a 25-page bill for consumer-facing AI agents, introducing a 'duty of loyalty' for agents to act on behalf of users and ensuring third-party agent access. The discussion draft is unlikely to pass this year.

Enterprise (2)

• California secured a deal with Anthropic for statewide deployment of Claude at 50% off, including free workforce training and technical support for all state and local governments. Governor Gavin Newsom emphasized AI should augment, not replace, human work.
• Anthropic renegotiated its pricing with Amazon, shifting from wholesale computing hours to token-based billing for Claude use, prompting Amazon to explore cost-saving alternatives like OpenAI or Nova models. Anthropic claims services are getting cheaper.

Big Tech (2)

• Meta implemented strict controls on engineers' use of Codex and Claude code in its Applied AI division since May to prevent rival model outputs from contaminating training data for internal frontier AI efforts. This presents a 'distillation trap' and legal exposure.
• Google capped Meta's Gemini usage and other large customers in March due to a compute crunch, causing Meta to prioritize its in-house Muse Spark model and pursue token efficiency. The Financial Times reports even major tech companies struggle to secure sufficient computing power.

AI Infrastructure (3)

• AWS hiked EC2 capacity block prices by 20% for Nvidia GPUs, indicating a shift from opportunistic spot GPU rentals to committed contract pricing for serious buyers. Spot H100 prices fell 40% from May, but SemiAnalysis notes contract prices continue rising.
• AI demand reignites the US power sector, with annual growth post-2024 at 150% of the historical average, reaching nine terawatt hours per month, compared to flat generation between 2008 and 2024.
• Value is shifting up the AI stack, with the app and model layer representing almost three times more AI revenue over the last year. Public companies reporting AI impact on earnings calls increased from 10% in early 2023 to 33% today.

Elections (3)

• Dave Portnoy is considering a New York City mayoral run, criticizing Zohran Mamdani's voter base as "elitist women." Krystal Ball noted Mamdani's favorability is rising among a diverse, working-class coalition, according to Michael Lang's analysis.
• A recent poll found Barack Obama most popular among Democratic voters at 54%, with Bernie Sanders and Zohran Mamdani also ranking high. Among Republicans, Marco Rubio led, while Tucker Carlson scored 28%.
• DSA candidates are making significant inroads in Democratic primaries, with Malt Kiros challenging Diana DeGette in Colorado and Peggy Flanagan leading Angie Craig in Minnesota.

Politics (3)

• Krystal Ball criticized public figures like Dave Portnoy and Bruce Blakeman for baselessly labeling critics of Israeli actions, such as Brad Lander and Graham Platner, as "Nazis" or "collaborators."
• Saagar Enjeti argued these "Nazi collaborator" smears are now ineffective and counterproductive for pro-Israel advocates, given evolving public opinion and a "shift of viewpoint" within the Democratic Party.
• Krystal Ball and Saagar Enjeti contend the Democratic establishment's loss of "credibility on electability" has fueled the rise of progressive candidates, who are successfully building broader, multiracial, working-class coalitions.

AI Infrastructure (2)

• Jeremy Grantham and the Bank for International Settlements (BIS) warn that massive spending on AI constitutes a "giant bubble" accumulating financial vulnerabilities that could trigger a global economic crash.
• Saagar Enjeti compared the AI industry's "China op" narrative, used to stigmatize data center opposition, to "RussiaGate," highlighting the dismissal of legitimate community concerns.

Markets (1)

• Saagar Enjeti predicted a simultaneous AI market crash and significant job displacement, comparing the situation to the calamitous but transformative railroad bubble, noting AI's unique goal of replacing human intellect.

Enterprise (1)

• Ford rehired hundreds of human engineers after AI quality checks failed to meet standards, tasking them with training AI for eventual replacement, a strategy Meta also uses.