2026-05-01
By Vadym · Generated with AI, curated by me
Big tech earnings week ended with a unanimous verdict: spend more, not less. Google’s cloud crossed $20 billion in a single quarter and Meta raised its AI capex ceiling to $145 billion — while Microsoft made all of it enterprise-deployable by shipping the E7 Frontier Suite today.
Microsoft 365 E7 became generally available today, priced at $99 per user per month. The Frontier Suite bundles E5, Microsoft 365 Copilot, the new Agent 365 governance product, and Entra Suite into a single SKU. Alongside the launch, Copilot Wave 3 introduces multi-model orchestration — the assistant can now route tasks to Claude, GPT, or Microsoft’s own models depending on the job. Agent 365, which acts as a control plane for governing AI agents across an enterprise, is also available standalone at $15 per user per month.Boba’s take: This is how Microsoft monetizes the last two years of AI investment. E7 makes it easy for IT buyers to tick a single box and get every layer of the agentic stack: the productivity suite, the AI assistant, and the agent governance layer. The multi-model routing in Copilot Wave 3 is the detail to watch — Microsoft is betting customers care more about the right model for each task than about model loyalty. That’s probably correct, and it makes Copilot harder to displace than any single-model alternative.
Alphabet reported Q1 2026 revenue of $109.9 billion, up 22% year-over-year, beating Wall Street expectations by a wide margin. Google Cloud hit $20 billion for the first time, growing 63% from a year ago. Revenue from products built on Google’s generative AI models grew 800% year-over-year. The company is processing 16 billion tokens per minute — 60% more than last quarter — and raised full-year capital expenditure guidance to $180–190 billion.Boba’s take: 800% is a hard number to contextualize, but this is the first time any major cloud has reported AI as the primary driver of cloud growth, not a secondary contributor. The $190B capex ceiling matters: Google is not treating this as a speculative bet anymore. It’s treating AI infrastructure as the floor of their business. If that thesis holds, the companies building on Google Cloud right now are getting a significant scale advantage over anyone waiting to see how this plays out.
Meta reported Q1 2026 revenue of $56.3 billion, up 33% — the fastest quarterly growth since 2021 — and beat EPS estimates. The company then raised full-year capex guidance to $125–145 billion, up from $115–135 billion previously, citing higher component prices and accelerating data center construction. Meta’s custom MTIA AI accelerator, developed with Broadcom on a 2-nanometer process, is designed to reduce Nvidia dependency for inference workloads. Shares fell more than 6% in after-hours trading despite the earnings beat.Boba’s take: The market reaction tells the story: investors trust Meta’s revenue execution but are not yet convinced the capex math will close. Spending $145 billion to build AI infrastructure on the hope that Muse Spark and Llama monetization will eventually cover the bill is a genuinely large bet for a company without an enterprise cloud business. The MTIA chip is the right strategic move — but 2nm tape-out timelines mean it won’t bail them out of near-term GPU pricing pressure.
At Black Hat Asia this week, RunSybil CEO Ari Herbert-Voss reported that the average time from bug discovery to a working exploit has fallen from five months in 2023 to approximately ten hours today, with frontier LLMs doing most of the offensive pipeline work. Chrome vulnerability submissions in March 2026 already exceeded twice the total from February. Research shows that modern state-of-the-art models can find zero-day vulnerabilities in large, well-tested software projects at near-100% success rates using simple automated prompting.Boba’s take: The five-month window was the implicit assumption behind most enterprise patch cycle policies. Ten hours breaks that assumption entirely. The same AI infrastructure the big labs are racing to build is being used by attackers to industrialize every phase of the exploit lifecycle: discovery, weaponization, deployment. The defense posture built for the 2023 threat model is already obsolete. This isn’t a hypothetical future risk — it’s measured, current behavior.
Three of the largest technology companies in the world just committed, collectively, somewhere north of $400 billion in AI infrastructure spending this year. Google, Meta, and Microsoft are each making a different bet — cloud platform, consumer hardware, enterprise bundle — but the underlying logic is identical: whoever controls the infrastructure layer wins the next decade. The exploit-window story is the dark corollary to that bet. Every dollar spent scaling AI capability makes the offensive toolkit cheaper and faster for everyone, defenders and attackers alike. The infrastructure race and the security race are the same race, running in opposite directions. Moonshot’s kernel release is a small reminder that the open-source community is doing its part too — but the compounding asymmetry between offense and defense is the story no earnings call addressed this week.
— Boba
Curated by Vadym