Fifty-five analysts cover Microsoft. Every one of them is looking at the same set of numbers, and the aggregate they produce is a Strong Buy with a $576.40 average target — 13% above the $509.22 close on 28 September 2026. Then look at what people betting actual money think. On Polymarket, for the trading week beginning 28 September, the contract on Microsoft touching $480 was priced at 74%, while the contract on it touching $525 sat at 58%. Read those together and the real-money crowd was assigning materially higher odds to a move below the market price than above it, in the same week, over the same horizon. That is not a view any of the 55 sell-side targets expresses, because the lowest of them is $440 on a twelve-month view and the framework has no way to say “probably down first, then up”. The most interesting thing about Microsoft right now is not the target. It is the gap between the two forecasting systems.
That gap exists because the sell side and the prediction market are pricing different questions, and the distinction is the whole argument. Analyst targets are twelve-month fair-value estimates built off a discounted earnings model — they answer “what is this worth?” Polymarket contracts are path-dependent touch bets over days — they answer “where does it go next?” For most stocks the two rhyme. For Microsoft in late 2026 they have come apart, and the reason is that the stock has just run 44% off its 25 June low of $352.83 without the twelve-month picture changing at all. Over the full year to 28 September, Microsoft is down 1.0%. The fundamentals improved enormously across that year; the share price did not. A market that has round-tripped a year while the business compounded is exactly the market where “worth” and “next” stop agreeing, and where using one to answer the other gets expensive.
Key facts
- MSFT closed at $509.22 on 28 September 2026 — down 1.0% over twelve months, but up 44% from the 25 June low of $352.83 — Nasdaq daily closes, retrieved 29 September 2026
- Consensus target $576.40 across 55 analysts; street high $870, street low $440; consensus rating Strong Buy — StockAnalysis, 29 September 2026
- FY2026 revenue $331bn, up 18%; Microsoft Cloud up 27%; Azure up 41% and past $100bn for the first time — Microsoft FY26 Q4
- Calendar-2026 capex guidance cut from roughly $190bn in April to about $175bn — Microsoft FY26 Q4 commentary, July 2026
- Nearly 90% of Microsoft Cloud revenue comes from customers outside the frontier AI labs, per CFO Amy Hood
- Microsoft 365 Copilot passed 30 million paid seats — Satya Nadella, FY26 Q4
- Polymarket, week of 28 September 2026: MSFT touching $480 at 74%, touching $525 at 58%
What is actually happening inside the business
Strip out the AI narrative and Microsoft’s fiscal 2026 was one of the strongest years a company of this size has ever printed. Revenue reached $331bn, up 18%. At a $3trn-plus scale, 18% growth is the number that should be startling, and it barely gets mentioned. Microsoft Cloud grew 27%. Azure grew 41% and crossed $100bn in annual revenue — a milestone no cloud business other than AWS has reached, and Azure got there growing faster than AWS did at the same scale.
The application layer is converting too. Microsoft 365 Copilot passed 30 million paid seats, which is the part of the AI story that actually carries software gross margin rather than datacentre depreciation. “We are advancing the frontier on the cost-to-outcome curve, ensuring every customer can turn tokens into business results,” chief executive Satya Nadella said alongside the FY26 Q4 results. That phrasing is doing deliberate work: cost-to-outcome, not capability. Microsoft has stopped selling benchmark scores and started selling unit economics, because unit economics is what a CFO signs off on.
So why is the stock flat over twelve months? Because the market spent the year arguing about the denominator. Microsoft told investors in April it expected roughly $190bn of calendar-2026 capital expenditure. The stock derated hard into late June — the $352.83 low is a 35% drawdown from the October 2025 high — on precisely that concern. Then, in July, the company did something the consensus AI-capex narrative does not allow for: it brought calendar-2026 guidance down, to about $175bn. A $15bn reduction.
Nobody rewarded the capex cut, and that is the tell
The dominant frame on hyperscaler AI spending is a ratchet: each quarter the number goes up, each quarter the free cash flow gets worse, each quarter the “when does this pay back” question gets louder. Microsoft broke the ratchet and the stock is still flat on the year. That deserves more attention than it has received, because it tells you what the market is actually worried about — and it is not the size of the number.
Two pieces of evidence point the same way. First, Amy Hood’s disclosure that nearly 90% of Microsoft Cloud revenue now comes from customers outside the frontier AI labs. That statistic is the direct rebuttal to the circularity bear case — the argument that hyperscaler AI revenue is partly vendors funding customers who buy the capacity back, inflating everyone’s numbers. If 90% of the revenue is ordinary enterprises running ordinary workloads, the circularity objection applies to a tenth of the base at most. Hood also told investors Microsoft should remain cash-flow positive through fiscal 2027, which distinguishes it sharply from peers now guiding to negative free cash flow.
Second, the OpenAI relationship was restructured, and the restructuring cut both ways. Microsoft holds roughly 27% of OpenAI, a stake worth something in the region of $135bn, and OpenAI has committed to $250bn of additional Azure spending. That is an enormous pre-funded revenue pipeline. It is also, precisely, the 10% of the cloud base that the circularity critique targets — and it is the reason Hood’s 90% figure had to be disclosed at all.
The wider financing pattern is worth holding alongside this. Compute commitments across the sector have started to carry equity components rather than plain contracts — Anthropic committed $11.6bn to Akamai and took a warrant struck at $111.33. When compute buyers start taking equity in compute sellers, the accounting question about who is really funding whom stops being academic. Microsoft’s 27% of OpenAI is the largest such arrangement in the industry, and the market is discounting it rather than capitalising it.
Reading the two forecasting systems against each other
Here is the arithmetic that makes the Polymarket-versus-analysts gap concrete rather than rhetorical.
Microsoft’s consensus target of $576.40 implies 13.2% upside. The street high of $870 implies 70.8%. The street low of $440 implies a 13.6% decline — and notably, a $440 target is a twelve-month valuation call, not a prediction about next week. Now take the same week’s Polymarket prices: touching $480 at 74%, touching $525 at 58%. A touch of $480 is 5.7% below spot. A touch of $525 is 3.1% above. The market was, in effect, pricing a near-certain wobble lower against a coin-flip-plus move higher, inside a single week, while the sell side maintained a Strong Buy on a twelve-month view.
Those two statements are not contradictory. They are answers to different questions, and a reader who treats the analyst target as a trading signal is quietly substituting one for the other. Prediction markets on single equities have grown quickly enough to matter here — Nvidia, Alphabet, Apple and Tesla now lead Polymarket’s stock-linked trading boom — and Microsoft sits in the same cluster of heavily traded names.
| The case for $870 | The case for $440 |
| Azure compounds past $140bn as the 41% growth rate persists into FY27 | Azure growth decelerates toward the high twenties as comparisons harden |
| Copilot seats scale from 30m toward 100m at software gross margin | Copilot seat growth stalls and per-seat pricing gets competed down |
| The $175bn capex number proves to be the peak; free cash flow inflects | Capex reverts upward and Hood’s cash-flow-positive FY27 guidance breaks |
| The 27% OpenAI stake gets marked as an asset rather than a liability | OpenAI’s $250bn Azure commitment is revealed as circular and gets discounted |
| 90% non-frontier-lab revenue proves the demand base is genuinely broad | Enterprise AI budgets consolidate and the long tail of Azure AI spend thins |
The $440 bear case deserves the same explicitness we gave the bull. At $440, Microsoft trades 13.6% below spot and roughly 19% below its 52-week high. That is not a crash — it is the stock revisiting where it traded in the spring of 2026, comfortably above the $352.83 June low. A $440 print does not require the AI thesis to break. It requires only that Azure decelerates by a few points while capex fails to come down further, which compresses the multiple without touching the narrative. That is the single most likely way this stock disappoints, and it is why the street low sits there rather than lower.
The regulatory and structural tension
Microsoft’s policy exposure has shifted shape over the past two years, and the current version is less about antitrust than about the architecture of its AI relationships.
The OpenAI restructuring is the centre of it. A 27% stake in the most scrutinised AI company in the world, combined with a $250bn compute commitment flowing back to Azure, is an arrangement regulators on both sides of the Atlantic have every incentive to examine. The UK’s Competition and Markets Authority has already tested whether the Microsoft–OpenAI relationship constitutes a relevant merger situation, and European scrutiny of cloud–model tie-ups has been persistent. The structural question is straightforward: if the largest customer of a cloud provider is also 27% owned by that provider, is the resulting revenue arm’s length? Microsoft’s answer is Hood’s 90% figure, which is a good answer but an indirect one.
The second tension is infrastructure, not law. Datacentre buildouts at the $175bn scale run into grid interconnection queues, local permitting and water-use politics, and those constraints do not respond to capital. A capex number is a statement of intent; delivered capacity is a statement of what the grid allowed. Any quarter where Azure growth disappoints because capacity arrived late rather than because demand softened will be read by the market as the latter, and that misreading is a real source of downside volatility.
The third is competitive and regulatory at once: bundling. Copilot’s 30 million seats were sold substantially into an installed Microsoft 365 base. That is an extraordinary distribution advantage, and it is the classic shape of the complaint European regulators brought against Teams bundling. A remedy that forced Copilot to be sold unbundled would not end the product, but it would change the seat-growth curve that the $870 bull case depends on.
What happens next
First, watch the capex line more carefully than the Azure line. Azure growth of 41% has become the headline, but the market has demonstrated across the past twelve months that it will pay for growth only if the spending trajectory is credible. The $190bn-to-$175bn revision is the most important thing Microsoft did all year and it was under-covered. If the next update holds at or below $175bn while Azure stays above 35%, that combination re-rates the stock more than a beat on either line alone.
Second, expect the free cash flow conversation to replace the capex conversation. Hood has committed to cash-flow positive in fiscal 2027. That is now a checkable claim, and it is the specific metric that separates Microsoft from peers who have stopped guiding to it. A confirmed positive-FCF year while spending $175bn would be the strongest argument available that this cycle is fundable from operations — and it is a comparison the market has not yet made explicit across the hyperscalers. The peer set is already fragmenting on exactly this axis, which is visible across the rest of the AI infrastructure complex from AMD down.
Third, treat near-term weakness as structurally likely rather than as thesis failure. The prediction market’s 74% on a $480 touch was, whatever happens to that particular contract, a reasonable read of a stock that had just run 44% in three months. Our working view is that $870 needs two more years of 40%-plus Azure growth to justify itself, which makes it a 2027–2028 number rather than a 2026 one, while $440 requires only a single soft quarter. The base case remains the boring one: a business compounding at 18% with a share price that spent a year going nowhere is not a broken stock, it is a coiled one.
Frequently asked questions
What is the Microsoft stock prediction for 2026?
Our bull case is $870, matching the street high, and our bear case is $440, matching the street low. The consensus of 55 analysts is $576.40, implying 13.2% upside from the $509.22 close on 28 September 2026. The consensus rating is Strong Buy.
Why is MSFT flat over twelve months when Azure grew 41%?
Because the market spent the year repricing the cost of that growth rather than the growth itself. Microsoft guided to roughly $190bn of calendar-2026 capex in April, derated 35% into the 25 June low of $352.83, then cut the guidance to about $175bn in July and recovered 44%. Net across the year: down 1.0%.
Did Microsoft cut its AI capex?
Yes. Calendar-2026 capital expenditure guidance came down from roughly $190bn in April to about $175bn by July. It is one of the few instances of a major hyperscaler reducing an AI spending forecast, and it received notably less coverage than the increases did.
What does Polymarket say about Microsoft stock?
For the week of 28 September 2026, Polymarket priced Microsoft touching $480 at 74% and touching $525 at 58%. Those are short-horizon path bets, not valuations, and they were more cautious near-term than any of the twelve-month analyst targets.
How much of Microsoft’s cloud revenue comes from AI labs?
Roughly 10%. CFO Amy Hood disclosed that nearly 90% of Microsoft Cloud revenue comes from customers outside the frontier AI labs, which is the company’s main defence against the argument that hyperscaler AI revenue is circular.
What is Microsoft’s stake in OpenAI worth?
Microsoft holds approximately 27% of OpenAI, a stake valued in the region of $135bn, alongside an OpenAI commitment to roughly $250bn of additional Azure spending. Both the stake and the commitment are subjects of ongoing competition-authority interest.
This article is for information only and is not investment advice. Price targets are scenario analysis, not forecasts.