The Omniscient Bulletin · 2026-05-20
The Omniscient Bulletin — May 20, 2026
Meta executes its layoffs, Nvidia prints $81.6B, and Google's Gemini 3.5 Flash goes live.
Nvidia just reported the most profitable quarter in semiconductor history - $81.6 billion in revenue, $58.3 billion in net income, guidance for $91 billion next quarter - all without selling a single chip into China. The number is so large it tends to flatten everything around it, but today's other stories are worth reading against that backdrop. Meta, the same day Nvidia printed those earnings, began notifying 8,000 employees that their jobs are gone, redirecting their salaries into a $125-145 billion capital spending plan aimed squarely at the infrastructure Nvidia builds. Google shipped a faster, cheaper model that beats its own flagship. And the White House moved to place itself inside the frontier model release pipeline. The common thread: the capital and the models are accelerating. The workforce absorbing that acceleration is not.
Compute
Nvidia Q1 FY2027: $81.6B revenue, Data Center up 92% year over year, dividend jumps 25x
Nvidia reported Q1 fiscal 2027 revenue of $81.6 billion on May 20, up 85% year over year and 20% sequentially. Data Center alone hit $75.2 billion, rising 92% from a year ago, driven by Blackwell 300 adoption. GAAP net income tripled to $58.3 billion. The board raised the quarterly dividend from $0.01 to $0.25 per share and added $80 billion to the buyback authorization. Q2 guidance of $91 billion assumes zero China Data Center compute revenue.
Labor
Meta's 8,000-person layoff wave begins, 6,000 open roles cancelled, 7,000 shifted into AI work
Meta started its first wave of cuts on May 20, notifying affected employees at 4am local time. The round eliminates roughly 10% of its global workforce. Among the first to go: integrity, cybersecurity, and content design teams. The company also cancelled 6,000 planned hires and is redeploying 7,000 existing staff into AI-related roles. Capital expenditure for the year is forecast at $125-145 billion, more than double 2025 spending, as Zuckerberg funds the Meta Superintelligence initiative.
Frontier
Gemini 3.5 Flash launches free to all users, outperforms Gemini 3.1 Pro on key benchmarks
Google released Gemini 3.5 Flash on May 20, the first model in its 3.5 series and now the default powering Google Search. The model scores 76.2% on Terminal-Bench 2.1 and 1,656 Elo on GDPval-AA, surpassing its prior flagship Gemini 3.1 Pro. It runs 4x faster on output tokens than competing frontier models and costs less than half what those competitors charge, priced at $1.50 per million input tokens. Google also introduced a new isolated agent runtime for single-call agent environment provisioning.
Industry
Google DeepMind pays $80-90M for Contextual AI talent in acqui-hire that draws antitrust attention
Google DeepMind finalized a deal to hire more than 20 researchers from Contextual AI, including CEO Douwe Kiela, for $80-90 million alongside a technology licensing agreement. The structure mirrors Microsoft's Inflection move and is already drawing regulatory attention: the deal is technically a talent-plus-license arrangement rather than an acquisition, an approach regulators in both the US and UK have flagged as a way to consolidate AI capability while sidestepping merger review thresholds.
Policy
White House moves to require early access to frontier AI models before public release, per Axios
The White House issued a directive on May 20 requiring AI developers to provide the federal government with early access to frontier models ahead of public release, according to Axios. The move reflects ongoing internal tension between factions favoring aggressive AI deployment and those pushing for safety review mechanisms. It extends a pattern of executive action on AI - this is at minimum the administration's fourth AI-related order since January 2025.