Bloomberg Tech
2026-06-05 · Hosted by Caroline Hyde, Ed Ludlow · Bloomberg / iHeartMedia
Executive Summary
A special edition broadcast live from the Bloomberg Tech event in San Francisco was dominated by Broadcom’s earnings disappointment, which wiped roughly $300 billion in market value as the company guided custom-silicon sales of $16 billion versus the street’s $17.2 billion expectation, triggering the worst day in chips since mid-May and the broader market’s worst session in 100 days. Against that backdrop, executives and investors debated whether AI is in a bubble — Cerebras CEO Andrew Feldman argued it is “the opposite of a bubble” with a $25 billion-plus backlog, while Databricks CEO Ali Ghodsi claimed AGI has effectively arrived. The event coincided with a historic capital-formation wave: SpaceX priced its IPO at $135/share seeking $75 billion at a ~$1.8 trillion valuation (pricing June 11), Anthropic filed confidentially Monday, OpenAI raised $122 billion, and Alphabet upsized an $84 billion equity offering. San Francisco Fed President Mary Daly said productivity gains from AI have not yet shown up in the data but expressed bullishness, and downplayed financial-stability concerns from rising markets.
Key Stories & Changes
1. Broadcom Earnings Trigger Chip Selloff
Broadcom guided current-period custom silicon (TPUs, ASICs) sales to $16 billion, missing the street’s $17.2 billion expectation
Stock fell as much as 14–15%, its biggest drop since January 2025, erasing roughly $300 billion in market cap
Company had added $270 billion in market capitalization in the run-up; it was “priced for perfection”
CEO Hock Tan spoke at the event; per Rosenblatt research, Tan cited visibility into 2028 that he did not have three months ago and remained bullish long-term
The broader market had its worst day in 100 days, off ~0.8%
2. The AI Bubble Debate
Cerebras CEO Andrew Feldman: “We’re in the opposite of a bubble” — backlog of more than $25 billion; “none of us, not AMD, not Nvidia” can keep up with demand
Databricks CEO Ali Ghodsi: contrarian view that “we actually already have AGI”; ~90% of audiences agree frontier models are smarter than most coworkers; the missing piece is context, which Databricks’ Genie product addresses
Ghodsi cited Novo Nordisk using Genie to shorten obesity-study analysis “from weeks down to a few minutes”
Ghodsi: in the next ~9 months to 2 years, “more software being written than in all history of mankind”
3. Historic IPO / Capital-Formation Wave
SpaceX: priced IPO at $135/share, seeking $75 billion, ~$1.8 trillion valuation; pricing June 11
Anthropic: filed confidentially Monday; Altimeter led its latest round
OpenAI: historic $122 billion raise; ~1 billion ChatGPT users, 5 million Codex users
Alphabet: upsized to $84 billion equity offering
Nvidia announced an $80 billion buyback; SK Hynix retired $8 billion of stock
4. Altimeter’s “Receiving vs. Spending CapEx” Framework
Altimeter partner framed AI as one of the largest capital-formation cycles in history
Key mental model: are you receiving the capex (compute, energy, memory, networking layers) or spending the capex (the labs)?
Most-common “long” theme among Stanford guest speakers: picks-and-shovels (compute), then energy as the ultimate bottleneck; most-common “short”: incumbents not innovating
Described XAI as a “Switzerland” business now selling compute, models, and coding products
5. Okta Raises Guidance on AI-Agent Identity Layer
Okta CEO Todd McKinnon: beat Q1 expectations last week and raised full-year guidance across the board
New product cycle is Okta for AI Agents — identity and secure connectivity layer for enterprise agents
McKinnon: “SaaS apocalypse” talk is “completely overblown”; expects Okta to have far more software engineers in five years
Models now exceed the ability to securely connect systems; connectivity (not model capability) is the barrier
6. Databricks on IPO Timing
Ghodsi: Databricks raised ~$20 billion; ~10,000 current employees (~14,000 including prior)
Will eventually go public to create a liquidity mechanism for employees, but “this is a terrible year” given how fast everything is moving
Not trying to time the market; not worried about private capital drying up
AVGO: Broadcom — -14% to -15% — Custom-silicon guide of $16B missed $17.2B expectation; ~$300B market cap lost
NVDA: Nvidia — Cited — $80B buyback announced; still can’t keep up with demand per Cerebras
GOOGL: Alphabet — Cited — Upsized equity offering to $84B
OKTA: Okta — Cited — Beat Q1, raised full-year guidance on AI-agent identity layer
Trends Identified
1. The AI Capital-Formation Supercycle
AI has shifted from a technology product cycle into one of the largest capital-formation cycles ever, with simultaneous mega-raises from SpaceX, Anthropic, OpenAI, and Alphabet plus buybacks from Nvidia and SK Hynix. The dividing line investors are drawing is whether a business receives the capex (compute, energy, memory, networking) or spends it (the labs) — a framework that reframes the entire AI value chain and explains why “picks and shovels” and energy dominate the bull case.
2. “Priced for Perfection” Fragility in AI Hardware
Broadcom’s selloff despite still-strong demand illustrates how thin the margin for error has become in AI semiconductors. A guide of $16 billion — an impressive number in absolute terms — was punished because expectations had run to $17.2 billion after a $270 billion run-up. The cyclicality of chips and the gap between demand and builder capacity remain the central tension.
3. The “SaaS Apocalypse” vs. Agents-as-Users Debate
Both Okta and Databricks pushed back hard on the narrative that AI commoditizes software, echoing Jensen Huang’s argument that agents themselves become the ultimate users of software. The bullish case holds that every layer of the stack is being reinvented and that demand for databases, identity, and connectivity grows as agents proliferate.
4. AI’s Real-Economy Impact Still Pending
Mary Daly repeatedly emphasized that transformative, economy-wide productivity gains from AI have not yet materialized, with ROI “still to be developed” and businesses pointing to “next year” as the litmus test. The enthusiasm and investment are real and broad-based (agriculture to machining to services), but the data has not yet caught up. —-
Sentiment Analysis
Overall Market Sentiment: Cautiously Optimistic
The mood blended exuberance over a historic IPO/AI wave with genuine anxiety that markets — particularly chips — have run too far, crystallized by Broadcom’s plunge.
Risk Factors Highlighted
AI hardware priced for perfection: Broadcom shows that even strong results can trigger massive selloffs when expectations are too high.
Chip cyclicality: Semiconductors remain a deeply cyclical industry vulnerable to demand swings.
Wall of IPO supply: Back-to-back trillion-dollar offerings (SpaceX, Anthropic, OpenAI) could strain market liquidity and absorb capital from other names.
Productivity gains unproven: AI ROI is “still to be developed”; failure to materialize next year would undercut the investment thesis.
Inflation from oil/food and the Iran conflict: Energy and fertilizer/food prices are the near-term inflation drivers with an uncertain war end.
Data-center financing / credit: Daly is watching how data centers are financed; “for some, it’s very worrying.”
Rate-path uncertainty: The Fed declined forward guidance, leaving 2026 rate cuts contingent on inflation and the war’s resolution.
AI labor disruption: Cautious hiring as businesses interrogate how much AI can do before adding headcount.
This episode was covered in today’s The Market Signal — 2026-06-05, a cross-source synthesis of multiple podcast reports.