The Erosion of Accountability: How Tech Giants Learned to Rewrite the Rules During a Boom
November 15, 2025
The numbers sound fine on the surface. Earnings growth is running at nearly 17% year over year, with 83% of companies beating expectations against a long-term average of 67%. The Nasdaq lost 2.3% on Thursday and managed to climb back into positive territory by 0.13% on Friday. Meta reported strong results but saw its stock plummet 11%, the worst day in three years. Bitcoin dipped below $95,000. Nothing catastrophic. Nothing that screams crisis.
But beneath the surface metrics, something more systemic is unraveling: the collective willingness to pretend that aggressive financial engineering is the same as real growth.
Consider what just happened. Meta narrowed its capital expenditure guidance to $70-72 billion from $66-72 billion, meaning management expects to spend even more aggressively on AI infrastructure. This spending has yet to generate meaningful revenue. The company's operating expenses jumped $7 billion year-over-year and nearly $20 billion in capital expense, resulting from intense spending on AI talent and infrastructure which has yet to bring in meaningful revenue. During the earnings call, when pressed on ROI, CEO Mark Zuckerberg offered reassurances about "latent opportunity" and future products "in the coming months." The market responded by erasing $20 billion in market cap in a single session.
This isn't unique to Meta. Oracle's stock has lost one-third of its value since September and is on pace for its worst month since 2011, having surged 36% in September on an OpenAI-related hype wave. Investors are questioning whether OpenAI can live up to its $300 billion commitment to Oracle over five years, with AI sentiment said to be "waning". Oracle's debt is expected to double to more than $290 billion by fiscal 2028, according to Morgan Stanley.
None of this is accidental. And the accounting matters.
The Depreciation Shell Game
In recent days, a voice from the 2008 financial crisis reemerged to flag something most institutional investors would rather ignore. Michael Burry has accused major technology companies of using aggressive accounting tactics to artificially boost their earnings amidst the AI boom, alleging that "hyperscalers" are understating depreciation expenses by estimating that chips will have a longer life cycle than is realistic.
Here's the mechanics: When tech giants like Microsoft, Meta, and Oracle build AI data centers, they buy tens of billions of dollars in GPUs, servers, and cooling systems. Normally, those assets lose value fast, cutting into profits. But recently, companies quietly extended how long they claim those machines last—from roughly three years to as many as six. That simple change lets them spread out their costs and report fatter earnings now.
The accusation isn't hypothetical. Meta's filings appear to corroborate the directionality of Burry's claim: until 2024, servers and network gear were depreciated over four to five years; effective January 2025, Meta said they would "extend the estimated useful lives" of "certain servers and network assets" to 5.5 years. Burry estimates that from 2026 through 2028, this accounting maneuver would understate depreciation by approximately $176 billion, artificially inflating reported earnings, with Meta's profits potentially overstated by roughly 21% and Oracle's by roughly 27% by 2028.
Now, is this technically illegal? No. Under GAAP, companies are given leeway in estimating depreciation. They have discretion. But discretion in accounting is a dangerous thing. It's the kind of discretion that precedes instability.
The irony cuts deeper when you consider the technological reality. The timing of these accounting changes makes little sense, as the pace of technological change accelerates—Nvidia now releases new chips every 12 to 18 months instead of every two years—meaning hardware becomes obsolete faster, not slower. Companies are claiming longer asset life precisely when those assets are depreciating faster.
The Circular Flow of Capital
There's another pattern worth examining. The entire AI spending cycle has become self-referential. A recent spree of circular AI deals among companies including Meta, Google, OpenAI, Nvidia, Oracle, and Broadcom has heightened fears that demand for AI could be overstated. Meta buys chips from Nvidia. OpenAI makes deals with Oracle and Microsoft. Those same companies buy more chips from Nvidia. Someone's capex becomes everyone else's revenue.
This isn't inherently wrong. But it's fragile. It creates the appearance of bottomless demand where the actual product utility remains speculative. When demand slows—when customers realize they've overbuilt, when customers realize the ROI doesn't materialize—the whole structure reverberates. There's no external demand cushion. Everyone is selling to each other.
JPMorgan announced it has officially completed the proof-of-concept for its USD-denominated deposit token, JPM Coin (JPMD), for its institutional clients. The token is available 24/7 with near-instant settlement on Base, an Ethereum Layer 2 blockchain network. Which is fine—stablecoin infrastructure is probably useful. But it also exemplifies the tech sector's constant motion toward "innovation" that solves no particular problem. There's optionality, liquidity, and the appearance of progress. But the underlying utility question never gets settled until there's real stress.
The December Gamble
As per CME Group's FedWatch tool, interest rates are predicted to have a 47.8% chance of a cut at the Dec. 9–-10 meeting, down from earlier levels near 52.2%, reflecting growing doubt inside the central bank. The Fed faces a genuine dilemma: inflation remains elevated, the labor market is holding firm, and yet the market expects easing.
This is the structural tension that should worry you. Tech companies have front-loaded massive capex in hopes that AI demand justifies it before interest rates rise further. If the Fed holds or hikes, the math becomes hostile. If the Fed cuts despite elevated inflation, central bank credibility faces a visible hit. Either way, something gives.
Historical Precedent
There's a pattern here that echoes through financial history. During the railroad boom of the 1870s, companies extended asset depreciation schedules to justify aggressive capital expenditure programs. During the dot-com bubble of the late 1990s, companies shifted to non-GAAP metrics to obscure deteriorating fundamentals. During the housing boom of the 2000s, banks repackaged risk until nobody understood what they owned.
Each time, the behavior was technically compliant. Each time, managers genuinely believed in the underlying opportunity. Each time, the accounting gimmicks existed not to commit fraud but to bridge the gap between the strategic vision and the financial reality—to buy time for the vision to manifest.
Sometimes it works. Most of the time, it doesn't. And by the time it fails, the damage is already embedded in the system.
The current situation with tech isn't a fraud. But it is an erosion of prudence. It's a collective decision to make accounting flexibility serve narrative rather than clarity. And it's happening during a period of genuine technical uncertainty about AI's economic impact.
What comes next depends on whether the capex actually generates returns, and whether the market—currently skeptical—eventually becomes convinced. But the skepticism itself is the real signal. When smart investors start asking hard questions about depreciation schedules, it means something deeper has shifted.
Trust in the narrative has fractured. And fractured narratives are how bull markets end.