TL;DR: The recent wave of mass layoffs in the US is not a sign of economic recession; it is the opening salvo in a brutal, physical war for AI dominance. This is a "Prosperity Depression" where companies are systematically liquidating human capital (Opex) to acquire compute capital (Capex), specifically GPUs. This article deconstructs the two forces driving this trend—tech giants firing to buy the "shovels" and traditional industry firing because they've struck "gold"—and explains why this is the macroeconomic proof that your unique, non-automatable human experience is the only viable career moat left.
I am James, CEO of Mercury Technology Solutions. October 31, 2025
In my last post, I argued that in the age of AI, your unique life experience is the only truly defensible moat. Today, we are seeing the stark, macroeconomic evidence of that thesis playing out in real-time.
Recent headlines are screaming about mass layoffs across the United States—48,000 at UPS, 30,000 at Amazon, 24,000 at Intel. Most analysts are diagnosing this as a symptom of a looming economic recession.
They are misdiagnosing the illness.
What we are witnessing is not a market contraction. It is a violent capital reallocation. This is not a cyclical downturn; it is a war. Human capital is being systematically liquidated to fund the acquisition of compute capital. The salaries of your employees are now in direct competition with the budget for NVIDIA's GPUs.
The "Prosperity Depression": Firing Engineers to Buy H100s
A traditional recession is defined by shrinking demand, which forces companies to cut costs. The situation in Silicon Valley is the exact opposite.
Amazon is not laying off 30,000 people because business is bad. On the contrary, business is booming. AWS has a backlog of $195 billion in orders, a 25% year-over-year increase. Customers are frantically placing orders, but AWS cannot deliver.
The bottleneck is singular and absolute: not enough GPUs.
The market is now ruthlessly punishing any hyperscaler that is slow to deliver compute. Amazon's only viable strategic move is to slash its operating expenditures (Opex)—namely, the salaries of its software engineers (SDEs)—and reallocate every available dollar to capital expenditures (Capex) to secure more NVIDIA GPUs.
Meta is operating on the exact same logic. On top of its regular layoffs, it has cut hundreds from its AI division. The reason is the same: a severe shortage of AI data center capacity. Their demand forecasts for compute have been revised upwards three times in the past year, and each time they have painfully underestimated the need.
This is the "Prosperity Depression": a state where a company's revenue and stock price are soaring, while its employees face Great Depression-level layoff anxiety. Your job is now competing for the same budget line as an H100 chip.
Two Paths to the Same Destination: Feeding the Compute Beast
This wave of layoffs tells two distinct but related stories. If tech giants like Amazon and Meta are firing people to afford the "shovels" (GPUs), then traditional giants like UPS, Nestlé, and Ford are firing people because they have already struck "gold" (AI-driven productivity).
These companies are laying off employees for the opposite reason: they have successfully deployed AI tools. Whether it's customer service automation, supply chain optimization, or generative design systems, the productivity gains are beginning to compound exponentially. They don't need to build their own massive GPU clusters; they "rent" inference compute from AWS or Azure. When the ROI math finally works, these companies are discovering they may no longer need the large human workforces of the past.
Both are feeding the same beast. Tech companies are buying the shovels; traditional companies are buying the gold that AI has dug. The result is the same: wealth is being concentrated from labor to compute on an unprecedented scale.
The New Value Chain: Semiconductors as the Ultimate Landlords
The biggest beneficiaries of this great reallocation are the "compute landlords" sitting in the middle: the semiconductor industry. NVIDIA, TSMC, and ASML are, for all intents and purposes, printing money. They are collecting a "compute tax" from both ends of the value chain. A new normal is emerging where the profit margins of semiconductor companies may soon eclipse those of internet companies.
This also explains the provocative argument that every software professional should probably own NVIDIA stock—not as an investment, but as a risk hedge. A hedge against the risk of being squeezed out of the value chain by the very GPUs you are being replaced to fund.
When Does the Bubble Burst? Watch the "50%" Adoption Rate
How long can this continue? The hyperscalers are desperately squeezing Opex, but at some point, there will be no more to cut. The next step will be to sacrifice cash flow and even take on debt (as Oracle has done) to acquire compute.
This is clearly a bubble, but history does not repeat itself perfectly. The key metric to watch is corporate AI adoption rate. Currently, across most industries, it's less than 10%. The fastest, most volatile phase of any technological revolution is the climb from 10% to 50% adoption. We are just entering that steep ascent.
When the dot-com bubble burst in March 2000, US internet penetration was around 52%. When this round of corporate AI adoption approaches 50%, the alarm bells will truly start ringing. The trigger for a collapse won't even require a reversal in demand, but merely a slowing of demand growth. The moment VCs and hyperscalers see that token demand is no longer growing exponentially year-over-year, they will cut orders without mercy.
Conclusion: This is Not a Recession. It is a Revaluation.
This brings us back to the central thesis from my last post. The macroeconomic forces at play are asking every professional a brutal and direct question: Is your value to this company greater than the value of the compute that could replace you?
If your work consists of repeatable, predictable tasks that can be learned from a vast public dataset, you are in direct competition with a machine. The only way to win this competition is to offer something that compute cannot.
Your defensible moat is no longer your skill in execution. It is your unique ability to synthesize disparate experiences, to build deep, human trust, to ask novel strategic questions, and to generate true, non-obvious creativity.
This is not a recession. It is a revaluation of human capital. And the only way to come out on top is to build a moat that no machine can cross.
Mercury Technology Solutions: Accelerate Digitality.
