Thesis Maps · Opened 22 Aug
Microsoft, Amazon, Alphabet, and Meta spent 165 billion on plant in one quarter. Token list prices can fall inside that mark. They do not unprint it.
Thesis map
A cheap token still arrives with a watt bill. Microsoft spent 35,802 million on plant against 17,079 million a year earlier.
Microsoft FY26 Q4 printed additions to property and equipment of 35,802 million dollars, against 17,079 million a year earlier. Amazon 54,208 million. Alphabet 44.9 billion. Meta 30,116 million. The four sum to 165 billion for one quarter. Token list prices can fall inside that plant mark. They do not unprint it.
The chain
Cheap token
A list price on a token can fall this quarter. A cooling tower cannot. Flatten a model name into a watt and you get a false headline.
Watt bill
Microsoft additions to property and equipment doubled. Those are cash-flow lines on the investor-relations page, not a broker table, not a 2026 guide.
Four-house plant
Amazon 54,208 million. Alphabet 44.9 billion. Microsoft 35,802 million. Meta 30,116 million. The four sum to 165 billion for one quarter.
Water bill
Goodyear 157th Avenue had no room above 500,000 gallons a day. Microsoft put 36 million toward a 90 million expansion and agreed later halls would be air-cooled. A token does not reprint that ceiling.
Long side
Owns the scarce object. Named filing, named plant, named queue.
MSFT
FY26 Q4 vs 17,079 a year earlier
Microsoft plant
35,802 million
Office, GitHub, LinkedIn, and Dynamics still sell seats. A licence is reprintable. 35,802 million against 17,079 million is a plant mark. One five-day window at 499.70 does not unprint it.
AMZN
Amazon plant
54,208 million
Amazon is the largest of the four cash lines this quarter. GW Ranch at up to 7.65 gigawatts behind the meter is a purchase agreement, not that cash line.
GOOGL
Alphabet plant
44.9 billion
Alphabet is the second cash line. MAREA, Dunant and Grace Hopper are the cables the cloud still comes ashore on. A token does not lay them.
META
Meta plant
30,116 million
Meta is the fourth cash line. Overview Energy named Meta farms for up to 1 gigawatt of night-time light, demonstration 2028. That is a 2028 beam, not this quarter plant.
Short side
Constrained by the chain, or a substitute that is not the product.
short
Model names
A list price is not a watt. A model name is not a cooling tower. Treating the API menu as the constraint hides 35,802 million of plant.
short
Office seat
The listed company still sells licences. The cash line is plant. Flattening them is how a software sleeve hides a builder.
short
API list
An API list that got cheaper is not a water permit. Goodyear already hit 500,000 gallons a day.
How they interconnect
Cheap tokenstill needsWatt bill
A cheap token still arrives with a watt bill. Microsoft plant doubled while the seat business kept selling licences.
Watt billis paid inFour-house plant
The four-house cash lines are the plant mark. 165 billion in one quarter is not a token menu.
Watt billstill needsWater bill
A watt bill still needs a water bill, or an air-cooled hall when the basin is full. Goodyear already wrote that ceiling.
Microsoft plantpaysWatt bill
Microsoft spent 35,802 million on plant against 17,079 million a year earlier. Azure is the reason that plant doubled.
Amazon plantpaysWatt bill
Amazon 54,208 million is the largest cash line. GW Ranch is a behind-the-meter offtake, not a substitute for that plant spend.
Alphabet plantpaysWatt bill
Alphabet 44.9 billion is plant. The cables are a different file. Both still sit under the token.
Meta plantpaysWatt bill
Meta 30,116 million is this-quarter plant. A 2028 Overview beam into Meta farms is not this cash line.
Microsoft plantoutrunsModel names
Plant spend outruns a model name. 35,802 million against 17,079 million is the observation.
Microsoft plantis notOffice seat
An Office seat is reprintable. The plant mark is not.
Amazon plantoutrunsAPI list
Amazon plant outruns an API list. A token that got cheaper does not cancel 54,208 million.
If you believe the thesis
Token prices can fall while the power bill and the water bill still rise. Azure is a power bid. The software seat is not the scarce object. The interconnect is. The megawatt is. 35,802 million against 17,079 million is a software house spending like a builder.
If the thesis is incorrect
Amazon or Microsoft print a quarter where cash plant spend falls while token volume still rises. Then the token was the constraint. Two consecutive quarters of aggregate hyperscaler capex guidance cut year on year would void the demand side, which is a different failure than a cheaper token.
A down week in the AI complex is supposed to be a verdict. The week of 17 to 21 August was a tape. Technology, the XLK fund, lost 3.5%. The Nasdaq 100 lost 2.45%. Friday's own session closed +0.11% and +0.33%. No margin printed. No capital-spending guide printed. The series that matter are quarterly. What did print, in July, is the size of the bill.
Amazon spent $54.2 billion on cash purchases and additions to property and equipment. Alphabet spent $44.9 billion. Microsoft spent $35.8 billion. Meta spent $30.1 billion. The four investor-relations and 10-Q cash-flow lines sum to $165 billion. Platformonomics prints $172 billion. Axis Intelligence prints $170.1 billion, equal to 99% of the four companies' operating cash flow in the same three months. Those third-party cuts mix leases and different bases. Combined 2026 guidance after the July prints sits around $735 to $750 billion.
Flagship API list prices, three official pages fetched 22 August 2026. OpenAI gpt-5.6-sol at $4 input and $20 output per million tokens on the standard Flagship table. The $5 and $30 pair is gpt-5.5 under 272k, not Sol. Anthropic Claude Opus 4.8 at $5 and $25. Opus 5 sits on the same list. Fable 5 at $10 and $50 is a different rung. Google Gemini 3.1 Pro Preview at $2 and $12 at or below 200k context, and $4 and $18 above it. DeepSeek V4 Pro is an order of magnitude cheaper on the published sheet. Cached input and batch cut those numbers again. The frontier is a ladder, not a monopoly price.
Token list prices can fall 60 to 75% while the cluster bill rises, because memory and power do not trade on the same sheet as output tokens. A cheaper completion is not a cheaper rack. Amazon moved a $200 billion cash figure to about $220 billion and said memory did part of the work. Microsoft and Meta had already cited component costs earlier in the year.
Andy Jassy's sentence is the one that belongs on this page. Even at $220 billion, Amazon will not have enough capacity to meet 2026 demand, and he believes the same will be true in 2027. Alphabet moved to $195 to $205 billion. Meta narrowed to $130 to $145 billion. Microsoft's calendar figure printed near $175 billion only because more data-centre leases are being classified as operating rather than finance leases. The company said the underlying build is unchanged, and the next quarter's capex, including that shift, would exceed $50 billion.
Amazon's AWS backlog of $496 billion, up from about $364 billion a quarter earlier, is the one number on the build that looks like an invoice. Jassy said Trainium2 is essentially sold out and much of Trainium4 is already reserved. Alphabet cannot point to an equivalent contracted figure. Meta cannot point to one at all, which is why the call spent time on inbound interest in selling compute externally. A backlog is not a gross margin. It is evidence that at least one of the four is not building into a void.
Do not start with a model demo. Do not treat a Friday in XLK as a spending cut. Do not use a cheaper token to explain a cheaper rack.
Axis Intelligence's 99% reading of capital spending against operating cash flow for the four names in the second quarter is the sentence the equity market has been refusing to write. Alphabet and Amazon each spent more on property and equipment last quarter than they generated in operating cash flow, on the cuts that have been circulating since the July calls. Meta's free cash flow after finance-lease principal was $784 million. Microsoft is the exception: operating cash flow still covered cash purchases, with free cash flow near $19.6 billion. A build that cannot be funded out of operations is a build that will be funded with debt. That does not end scarcity. It changes who gets paid if the cost of capital rises: the lessor, the bondholder, the memory vendor, before the application vendor does.
Picks and shovels, accelerators, memory, networking, shells, power, have captured nearly all the equity value so far. Their pricing power is a queue. High-bandwidth memory and turbine slots are allocated, not shopped. When unit volumes double and vendor gross margins hold, scarcity has not ended. When volumes double and margins fade, the supply response arrived. A week in XLK does not settle that.
The model layer is the one the market watches. Benchmarks are the tell it watches. They are the wrong tell. When the frontier is a commodity, the margin goes to whoever has distribution, not whoever has the best benchmark.
The application layer is where abundance either shows up as the vendor's margin or as the customer's saving. ICONIQ's 2026 State of AI snapshot put inference at roughly 23 cents of every AI-product revenue dollar at scaling-stage business customers, and average AI-product gross margin at 52%, better than 41% in 2024 and 45% in 2025, and still a floor that would have been a crisis print for a 2019 software underwriting model. Snowflake's product GAAP gross margin is 71%, non-GAAP 75%, down from a world that talked about 78%. ServiceNow's fiscal 2026 GAAP subscription margin guide is 75%, and the company has named AI and hyperscaler partnerships in the same sentence.
Private equity's old software move was to buy the 80% margin and cut the costs that sit above it. Inference sits below it. You cannot fire a token. Caching and distillation are real. They do not restore the underwriting law. The dollars rotate from engineers and support into cost of goods. A roll-up that treats that as a temporary tax will cut the product when it cuts the bill.
Hosting cost as a share of revenue is the next quarter's tell. Several first-quarter 2026 management discussions already isolate 4 to 9% of revenue as inference. The names that have not started that sentence are the names whose 80% margin is a timing difference. Auditors will force the classification: inference required to deliver the subscribed feature is cost of goods, not research.
XLK down 3.5% with the Nasdaq is the short truth. It will be quoted as a refutation of a compounding build. It is not. A refutation would look like this: the four companies' cash purchases go flat to down year on year for two quarters with no corresponding rise in disclosed AI revenue; or flagship list prices stop falling for four quarters; or vendor gross margins hold while unit volumes double. A Friday session is none of those.
NVIDIA on a Wednesday night is an input to whether the build is still sold out. Earnings per share and the data-centre number are not a software gross-margin event. Do not write a sentence that uses one to explain the other.
Power is the one layer where scarcity is not ending on a software clock. That argument lives on the power page. A campus that cannot point to a turbine is not an application-margin event.
IT-services revenue per employee is the clean public series for labour substitution. Rising with flat headcount is the substitution arriving. Services firms can look healthy on revenue while the headcount line is the whole story. Official productivity statistics are last to know. You see the computers everywhere except in the productivity numbers. That line is a decade-scale series. It will not settle this page inside a quarter.
Amazon's backlog is a fact about one invoice book. Snowflake's 71% is a fact about one product margin.