The Big Short, Part II?

When it comes to AI, ignore the stock charts — the real story may be in the amortization schedules. Pundits who spent last week watching the red arrows on the Nasdaq (down 2.7% for the week) may have missed the real story: The sell-off was noise, when the signal is in the accounting.

(Tap or click the headline above to read this story with all of the links to our background as well as external sources.)

What we’re seeing now is the real-time collision between two fundamental views of the AI economy: The “deployment phase” optimists who see a productivity boom, and “capex realists” (think: Michael Burry) who see (at best) a hole in the balance sheet that revenue cannot (yet?) fill.

The bulls see a deployment dividend on the horizon. The New York Times has recently argued that the “hype cycle” is coming to an end and the “deployment cycle” has begun. The argument: AI is no longer a parlor trick — it’s starting to show up in GDP and output-per-hour data. Bulls claim we’re at the start of a structural shift comparable to electrification or the internet — but happening at 2x the speed.

• The evidence: The cost of adoption is stabilizing while the efficiency dividend widens, suggesting the path to profitability is clearer than the market thinks.

The bears see a capex treadmill — and maybe some hinky accounting. There’s a flaw in the Internet 2.0 analogy, and it’s in the physics of the infrastructure. When the Dotcom bubble burst, it left behind mns of km of “dark fiber” — assets with a >20-year useful life that powered the internet for decades at near-zero incremental cost.

The catch: Today’s AI boom is built on silicon, not glass — and GPUs have short lifespans. Unlike fiber, GPUs are depreciating assets with “frontier lifespans” of roughly 3-4 years before energy efficiency gains make them economically obsolete.

The accounting hijinks: Hyperscalers like Microsoft and Google have quietly extended their server depreciation schedules to 5-6 years to protect current earnings. If the hardware becomes obsolete in three years, those earnings are a mirage.

The “re-buy” problem: As Sequoia’s David Cahn noted last year, the industry isn’t just building a railroad — it’s building a railroad where the tracks dissolve every 48 months. To stay in the game, Big Tech must re-spend its capex budget in perpetuity.

Famed contrarian Michael Burry (of The Big Short fame) has reportedly placed a USD 1.1bn wager against Nvidia and Palantir, warning that the sector is sleepwalking into a crash. His argument mirrors the depreciation thesis: He accuses hyperscalers of inflating earnings by artificially extending the “useful life” of their servers — a “depreciation trick” that boosts profits on paper while the hardware rots in reality.

The bottom line: If the Times is right, the revenue arrives just in time to pay for the next generation of chips. If it’s wrong, we aren’t looking at a bubble, but at the most expensive depreciation write-down in corporate history.

MARKETS THIS MORNING-

Asian markets are mixed in early trading this morning, with both the Hang Seng and Kospi up 1.1% and the Shanghai Composite down 0.2%. Meanwhile, Japan’s Nikkei is closed in observance of Labor Thanksgiving Day.

THE CLOSING BELL-

The EGX30 rose 0.4% at yesterday’s close on turnover of EGP 4.7 bn (5.2% below the 90-day average). International investors were the sole net sellers. The index is up 36.0% YTD.

In the green: Telecom Egypt (+3.8%), Madinet Masr (+2.3%), and Misr Cement (+2.3%).

In the red: Oriental Weavers (-2.2%), Credit Agricole (-1.7%), and ADIB (-1.6%).