
The 3.6% Productivity Quarter Is Too Thin to Price an AI Boom
America's output per hour is rising. The annual trend is less impressive, and the companies financing AI infrastructure still have to earn their returns.
In the third quarter of 2026, America's nonfarm businesses produced 3.7% more output at an annualized rate while labor hours rose just 0.1%. We would not pay a premium for AI infrastructure on the strength of one good productivity quarter. The preliminary BLS figures, as of 5 November 2026, put productivity growth at 3.6%. A premium needs incremental income that covers new capital and depreciation at an acceptable share price. Insufficient evidence for a premium is not a claim that the shares are overpriced.
The Headline Trend Gap is the risk of projecting the 3.6% quarterly annualized productivity reading into lasting growth when the year over year reading is 2.1%. Year over year growth fell from 2.8% in 2026 Q1 to 2.3% in Q2 and 2.1% in Q3, a 0.6 percentage point decline from 2025 Q3's 2.7%. The quarterly annualized rate rose from 2.1% in 2026 Q2 to 3.6%. An investor treating 3.6% as the new normal would pay for supplier earnings that the year over year 2.1% does not support. Microsoft still has to fill its datacenters at prices that earn money on the investment.
Federal Reserve Chairman Kevin Warsh's 16 November 2026 remarks envisaged faster productivity benefiting workers and investors, growth with displacement, or investment disappointment followed by falling valuations. He called these "illustrative scenarios, not forecasts." Warsh assigned neither investment dollars nor probabilities. An efficient economy can still leave its equipment investors with poor returns.
Microsoft's datacenters account for much of the $194.060bn in purchase commitments listed in its FY2026 Form 10-K. Of that, $169.008bn is short term and $25.052bn long term. Shareholders are paying for capacity before it starts paying them.
Microsoft generated $182.935bn in operating cash and spent $115.948bn on property and equipment in FY2026. $182.935bn of operating cash over $115.948bn in equipment spending gives 1.58 times coverage. In a flat cash scenario, the same $182.935bn less $115.948bn leaves $66.987bn before other uses. Microsoft has the cash to build. The Value Capture Test is whether it keeps enough of its customers' savings to earn more than its required return on new capital.
Microsoft's FY27 Q1 release puts Agents and Infra operating income at $40.558bn, up from $32.718bn. Revenue rose from $61.672bn to $76.795bn. You are buying the next increment of earnings, not the last one.
In FY27 Q1, cash additions to property and equipment climbed from $19.394bn to $39.367bn. The extra $19.973bn came alongside $7.840bn in additional Agents and Infra operating income. That is 39 cents of added quarterly income per extra capex dollar, or $7.840bn over $19.973bn. Shareholders want the profit to follow the spending.
Agents and Infra earned $7.840bn more operating income on $15.123bn more revenue, from $76.795bn less $61.672bn. That gives a 51.84% incremental operating margin after depreciation. Depreciation, amortization and other rose from $8.147bn to $11.744bn. Net property and equipment went from $313.076bn on 30 June 2026 to $346.844bn on 30 September 2026. These are company wide capital outlays and broad segment earnings, not an AI specific cash return. Shareholders need that margin to survive the expanding asset base.
Microsoft's FY27 Q1 release shows $677bn in commercial remaining performance obligations, up 73%. Microsoft Cloud revenue was $63.9bn, up 30%, and Azure grew 48%. Those contracts give Microsoft customers for its capacity. Shareholders collect the margin left after supplying them.
The BLS numbers show unit labor costs falling 0.7% at an annualized rate in 2026 Q3. Hourly compensation rose 2.9%, and real hourly compensation rose 1.1%. These are sector averages, not individual raises. You cannot assume the supplier keeps what enterprises, workers or customers save.
Mary C. Daly wrote on 23 February 2026 that "most macro-studies of productivity growth find limited evidence of a significant AI effect". The BLS year over year reading is 2.1%, and the amount attributable to AI remains unknown. A supplier's shareholder needs the difference.
The strongest bull puts Microsoft's $182.935bn in FY2026 operating cash beside its $677bn in commercial remaining performance obligations. That is funding and contracted demand in the same business. Customers filling the datacenters at firm prices can make early construction pay off. A shareholder who waits for every return to appear has missed that opportunity.
OECD authors Filippucci and colleagues model 0.1 to 1 percentage point of annual labor productivity growth from AI over the next decade in their central medium adoption scenario. That gives the bull a long horizon for adoption. It measures national productivity, however, rather than Microsoft's income. The Value Capture Test still turns on the money left after Microsoft serves those customers and pays for the machines.
Microsoft is today where the telecom builders stood in 1996, spending first and waiting for the cash.
The BLS historical analysis records annual readings of 1.1% in 1995, 3.8% in 1999, 4.3% in 2002 and 2.2% in 2005. These average 2.84% annually from 1995 to 2005. Yet the Richmond Fed analysis records communications equipment investment rising from about $62bn annually in Q1 1996 to over $135bn in Q4 2000, in constant 1996 dollars. They write, "Corporate profits for the communications industry started on a rapid downward trend after 1996". The Nasdaq telecom index peaked at 1,230.06 on 10 March 2000 and fell 62% by year end amid excess fiber capacity. The match is that useful infrastructure can leave builders with poor returns. The break is documented long haul fiber overcapacity; Microsoft's cited filings establish no AI capacity overhang. Aggregate productivity persisted while suppliers lost profits and valuations; the data do not attribute the entire gain to IT.
BLS Table 2 has revised quarterly annualized productivity growth of 4.5% in 2025 Q3, followed by 1.2% in Q4. The preliminary 3.6% for 2026 Q3 follows a revised 2.1% for Q2. A shareholder pricing 3.6% as permanent should reckon with the earlier shift from 4.5% to 1.2% within one quarter.
The BLS wage and unit cost figures show lower labor costs per unit of output alongside higher average real hourly compensation. You may own a business producing more per hour while its infrastructure supplier struggles to price the service.
The Headline Trend Gap warns against buying years of growth from one good quarter. The Value Capture Test asks whether Microsoft gets paid enough for building the capacity. We would be wrong to pass on the premium if Microsoft's fiscal second quarter 2027 release shows an incremental year over year operating margin in Agents and Infra of at least 51.84% while cash additions to property and equipment again rise year over year, demonstrating incremental profitability after depreciation during the capital build. That margin test gives neither a return on invested capital nor a share price. Until Microsoft's new capacity produces enough cash to cover the investment and justify the share price, we pass on the premium.