Who gets paid first in the AI boom?
Chip vendors book the boom now. Software companies can book payroll savings now. The productivity gain can remain a forecast.
The AI boom pays infrastructure vendors when they ship. A software company can take the same boom as lower payroll before its AI layer earns a dollar of new revenue.
SaaS was not starving
For the quarter ended April 30, GitLab reported $264.2 million in revenue, up 23% year over year. Its GAAP gross margin was 86%. It generated $149.2 million in operating cash flow and $146.7 million in adjusted free cash flow.
Eleven days after that quarter closed, GitLab announced a workforce reduction and reaffirmed full-year guidance.
NVIDIA reported $81.6 billion in revenue and a 74.9% GAAP gross margin, then returned about $20 billion to shareholders. TSMC reported a 66.2% gross margin. SK hynix reported a 72% operating margin.
The margins are all real, but they sit on different income-statement lines and answer different questions. GitLab’s workforce reduction was an allocation decision, not evidence that software had no margin.
GitLab was not missing margin. It was missing proof that the new AI layer deserved a separate invoice.
The AI boom is a payment queue
The chip vendor recognizes revenue after the infrastructure is delivered. It does not wait for the downstream application to earn the money back. The cloud owner takes utilization risk. The application company takes willingness-to-pay risk.
GitLab is not literally sending severance money to SK hynix. The connection is capital timing. The infrastructure seller books the boom now. The software company can book payroll savings now. The productivity gain can remain a forecast.
GitLab’s announcement says the restructuring and strategy are “related but independent”. Two paragraphs earlier, it says AI agents will automate reviews, approvals, and handoffs, followed by role “right-sizing.”
One HN commenter called the decision “in spite of LLMs, not because of them”. I think that is closer to the financials than GitLab’s framing, with one addition. AI does not need to replace the jobs. It only needs to make the reduction legible as strategy.
The thread was submitted by an account named AnonGitLabEmpl. Its first comment said employees could “have some anxiety until then. As a treat.” Whether or not the handle is literal, the uncertainty landed on employees immediately.
My bet
For the next two years, the bottleneck vendors keep winning. Generic AI wrappers get crushed. Established SaaS companies keep their gross margins by charging separately for agent work and defending operating margins with fewer people. GitLab already says subscriptions will stay and consumption pricing will cover the work agents do.
I also expect most AI-era headcount cuts to survive when the productivity numbers disappoint. The payroll saving has an effective date. The promised output is still in a slide deck.
HBM orders will eventually be cut when the cycle turns. I do not expect the headcount to come back.