I've been thinking about a question lately: the scale of AI's capex has already proven a supply-side explosion, but is this the "start of a new cycle" or "bubble acceleration"? The answer depends on an overlooked variable—the ultimate destination of the money saved by AI.

I categorize the flow of AI efficiency dividends into three paths:
Path A: Capitalists exclusively benefit (profits → buybacks → shareholders' pockets)
Path B: Reinvestment into new businesses and markets (J-curve initiation)
Path C: Lower service prices, benefiting consumers. Whether the Kondratiev wave can truly rebound depends on whether it follows A or C. Choosing B is a bubble.

Currently, the strongest signal is Path A. Big Tech profits hit new highs + unprecedented buyback scale.

White-collar hiring is cooling. SaaS sector valuations have been compressed from 15-20x P/S to 7-8x, with the reason being "AI reduces the need for fewer employees and less software."

A February study by NBER hits even harder: 90% of companies say AI has no impact on productivity, yet CEOs still predict AI will boost productivity by 1.4%.

There's a vast gap between expectations and reality.

The most ironic is the SaaS sector. ServiceNow Q4 subscription revenue grew 21%, cRPO increased 25%, performance was perfect—yet the stock price was halved.

The market isn't saying "you're doing poorly," but rather "your business model itself is devaluing."

Every dollar invested in AI infrastructure is one dollar not flowing into Salesforce seats. This isn't an individual stock issue; it's a systemic reset of valuation anchors.

The mobile internet also experienced a switch from Path A to Path B.

2010-2012: Smartphone explosion, but only hardware companies profited; application layer was still burning cash, traditional industries were disrupted, but new employment hadn't emerged yet.

2012-2013: iPhone 5 + 4G adoption + Uber/Airbnb explosion = emergence of native business models = real demand-side takeoff.

AI is roughly at the 2011 position now. The validation period is 2027-2028.

No one currently thinks AI is useless; rather, "AI is too useful but only for capitalists."

Supply-side explosion + demand-side suppression = bubble characteristics, not new cycle features.

Only when a large number of native business models emerge with the belief that "AI cannot exist without" will we see a true Kondratiev wave rebound.

Until then, maintain reverence.