Almost everyone reading Micron’s chart draws the wrong lesson from it. The stock went up roughly 8.4 times from its 52-week closing low of $115.79 to $971.66, so the instinct is to call it a bubble, or to say the move is over. Check the multiple and that reading collapses. Micron trades at 21.9 times earnings — a perfectly ordinary number for a semiconductor company. It did not re-rate at all. Trailing net income grew 710.7% to $50.47bn, and the share price simply chased it. That distinction is the whole of what follows, because it reframes the question people ask about Nebius (NASDAQ: NBIS) at $277.68. For Nebius to be the next Micron, it does not need investors to pay a higher multiple. It needs the opposite: it needs to de-rate into its own growth. On its own year-end guidance, it does exactly that — and Micron’s current multiple, applied to that guidance, prints $356 a share.
The insight: the bull case is a falling multiple, not a rising one
Here is the arithmetic that almost no one runs. Nebius carries a $76.11bn market capitalisation on 274.10m shares against trailing twelve-month revenue of $1.36bn. That is 56.0 times sales, a number that looks indefensible and gets the stock called a bubble on a daily basis. But revenue grew 506.9% over that period, so the trailing figure describes a company that no longer exists. Measured against the $3.0bn of annualised recurring revenue Nebius had actually reached by 30 June, the multiple is 25.4 times. Measured against the $7bn–$9bn year-end run-rate the company has guided to, the midpoint gives 9.5 times.
Micron trades at 12.2 times trailing sales today. So Nebius, at an unchanged share price, passes through Micron’s current valuation on the way down and ends up cheaper than it — purely by hitting its own guidance. Invert that and you get the headline number: apply Micron’s 12.2 times to Nebius’s $8bn guided run-rate and you get a $97.5bn market capitalisation, or roughly $356 a share, 28.1% above the current price. The $7bn low end gives $311; the $9bn high end gives $400.
That $356 is my arithmetic, not an analyst’s target, and it is a scenario rather than a forecast — it assumes Nebius hits guidance and that the market is willing to pay a memory manufacturer’s multiple for a compute landlord, neither of which is guaranteed. But it is the honest way to express the thesis, and it explains why the stock keeps confounding people who anchor on trailing numbers. Having watched this same confusion play out across the memory complex all year — through Micron’s own bull and bear case and the peak-cycle fear that dominated the NAND narrative — the pattern is consistent. Shortage-driven businesses look expensive on trailing data right up until the earnings land, and then they look cheap in hindsight.
Key facts
- NBIS last close $277.68, up 8.88% on the day; 52-week closing range $64.06 to $286.69 — 14 August 2026 (StockAnalysis)
- Street consensus $226 — below the current price, across 18 analysts; high $410, low $120 (MarketBeat, August 2026)
- Q2 2026 revenue $582m, up 454%; AI cloud revenue $575m, up 514%; adjusted EBITDA $236m at a 41% margin, against a $21m loss a year earlier — Nebius, 12 August 2026
- ARR $3.0bn at 30 June, up 56% from $1.9bn in March; year-end run-rate guidance $7bn–$9bn
- Customer prepayments cover 50%–60% of associated capex on four Q2 contracts averaging over $1bn each — Nebius Q2 2026
- Micron: 16 strategic customer agreements, ~$100bn minimum contracted revenue, covering ~20% of DRAM volume and a third of NAND volume through calendar 2030 — Micron fiscal Q3 2026
- Micron trades at 21.9x earnings after an 8.4x move, because net income grew 710.7% to $50.47bn (StockAnalysis)
What Micron actually did — and it was not riding a price spike
Micron’s fiscal third quarter of 2026 produced $41.5bn of revenue, up 74% sequentially and 346% year on year, at a company-record 84.9% gross margin. Those are the numbers everyone quotes. The number that explains the durability is different: 16 strategic customer agreements representing approximately $100bn in minimum contracted revenue, covering roughly 20% of DRAM volume and about a third of NAND volume through calendar 2030.
That is the structural change. A memory manufacturer’s historic problem was never demand — it was that demand arrived at prices set by a brutal spot market, so good years were unbankable and the equity never earned a durable multiple. By pre-selling a fifth of DRAM volume years forward at contracted minimums, Micron converted the least predictable part of the business into something closer to an infrastructure contract. CEO Sanjay Mehrotra called the quarter “exceptional,” with results that “exceeded the high end of guidance across all metrics.” HBM4 has already shipped over $1bn of revenue.
The tell that this is a genuine shortage rather than a hype cycle is the direction the money flows. In a normal market, a supplier funds its own capacity and hopes customers show up. In a real shortage, customers pay in advance to reserve supply, because the risk of not having it exceeds the cost of pre-committing. Micron’s $100bn of minimum contracted revenue is that signature at scale.
Nebius is showing the same signature, one stage earlier
Nebius closed four AI cloud contracts in Q2, each averaging more than $1bn, with Reflection, Cohere, a US “neolab” and a US quantitative trading firm. The terms run one to three years at a revenue yield of $20m–$25m per megawatt. The critical detail is the financing: customer prepayments cover 50% to 60% of the associated capital expenditure. Nebius’s customers are funding roughly half of the buildout that serves them, in advance. That is Micron’s prepayment dynamic, arriving in compute.
The pricing tells the same story from the other side. While long contracts price at $20m–$25m per megawatt, Nebius sells shorter capacity — up to six months — at $40m–$50m per megawatt, occasionally higher, to customers who need dedicated clusters for time-sensitive training runs. Spot is roughly double contract. A supplier that can charge twice as much for immediacy is not operating in a competitive commodity market. Founder and CEO Arkady Volozh put the resulting position bluntly: “We choose when to sell, to whom we sell, and on what terms, and how we finance everything.”
The operating leverage is already visible rather than promised. Adjusted EBITDA swung to $236m, a 41% margin, from a $21m loss a year earlier and 32% in Q1. Revenue reached $582m in the quarter, up 454%, with the AI cloud segment at $575m and 98% of the group. Capex ran at roughly $5.7bn in the quarter against full-year guidance of $20bn–$25bn, targeting 5 GW of connected power by year end. Our full breakdown of the Q2 numbers covers the capex quarter in detail.
Nvidia’s position is the other structural signal. In a Schedule 13G filed on 13 July 2026, Nvidia disclosed beneficial ownership of 22,256,412 shares — 1,190,476 held directly plus 21,065,936 underlying a pre-funded warrant acquired on 11 March — for 9.3% of the company, stemming from a $2bn strategic investment. Choosing a 13G over a 13D signals passive intent. Nvidia allocating both capital and, implicitly, supply priority to a customer is the closest thing to a qualification decision this industry produces.
Where the analogy breaks, and it breaks hard
Any honest version of this thesis has to state the disanalogy plainly, because it is severe. Micron manufactures a physically scarce product that is extraordinarily difficult to make. That is why it earns an 84.9% gross margin and why its moat compounds. Nebius rents out someone else’s chips. It must buy GPUs from Nvidia — its own 9.3% shareholder — at whatever Nvidia charges, and its economics are bounded by that input cost forever. A 40% adjusted EBITDA margin is a good business; it is not an 84.9% gross margin, and no amount of scale closes that gap.
The funding asymmetry is just as stark. Micron self-funds its capacity out of $50.47bn of trailing net income. Nebius is guiding to $20bn–$25bn of capex against a $76.11bn market capitalisation and trailing net income of $42.40m — essentially zero. Prepayments cover half of it, which leaves roughly $10bn a year to be financed from somewhere, and that somewhere is debt or equity. The comparison that matters here is CoreWeave, whose $104.2bn backlog sits alongside a debt load that dominates its story. Backlog is not cash, and neoclouds are financing businesses wearing technology clothing.
Then there is the competitive question nobody has answered. GPU rental has no obvious technical moat. If capacity catches up with demand, the $40m–$50m per megawatt spot pricing is the first thing to go, and the contracted book becomes a floor rather than a springboard. Micron’s shortage is enforced by physics and by a three-player oligopoly. Nebius’s is enforced by a temporary imbalance between Nvidia’s output and everyone’s ambition — a condition with no guarantee of permanence.
The street is not on board, and that is the live tension
The most striking fact in the data is that Nebius trades above where analysts think it should. Across 18 analysts the consensus target is $226, roughly 19% below the $277.68 close. The dispersion is extraordinary: Northland’s Nehal Chokshi raised to $410 on 20 July, Robert W. Baird went to $340 on 13 August and Citi to $324 on 14 August, while DA Davidson sits at $175 on a Neutral and Morgan Stanley’s Josh Baer carries $144 at Equal Weight. A high target 3.4 times the low one is not a disagreement about next quarter. It is a disagreement about whether this is infrastructure or a rental business in a cyclical upswing.
The near-term regulatory-style overhang is not a regulator at all but a lock-up. Nvidia’s contractual restrictions prevent it from exercising the warrant or selling the underlying shares before 11 September 2026. We covered what that date means for the stock when the stake was disclosed. A 9.3% holder becoming free to sell is a mechanical supply event regardless of intent, and it lands inside the next month. Anyone underwriting the $356 case should expect that date to be noisy.
What happens next
Prediction one: the year-end run-rate number is the entire thesis, and it is checkable. Nebius has guided to $7bn–$9bn of annualised run-rate revenue by year end, from $3.0bn ARR in June. That is roughly a tripling in six months, and it is the single input that drives every valuation conclusion here. Hitting the midpoint validates the $356 arithmetic. Landing at $5bn does not just miss — it resets the multiple to 15.2 times, above Micron’s, and the entire “cheaper than Micron” argument disappears.
Prediction two: the prepayment percentage matters more than the contract count. Watch whether prepayments stay at 50%–60% of associated capex on new deals. If that ratio holds or rises, the shortage is real and customers are still bidding for certainty. If it drifts down, it means Nebius is having to fund its own growth to win business, which is the first sign the market is normalising — and it would show up long before pricing cracks.
Prediction three: 11 September resolves an overhang in one direction or the other. Either Nvidia’s lock-up lapses without a sale, which reads as an endorsement and removes a discount, or paper starts moving. Given the stock already trades 23% above consensus, that date is the most likely near-term source of a sharp move in either direction.
The uncomfortable conclusion is that both the bulls and the bears are anchoring on the wrong number. Bears point at 56 times trailing sales and call it absurd; bulls point at 454% growth and call it inevitable. The number that decides it is the year-end run-rate, because that is what converts an expensive-looking stock into a cheap-looking one without the price doing anything at all. Micron’s investors learned that lesson the slow way, watching a stock they thought had run too far keep pace with earnings that ran further. Nebius is at the stage Micron was at before the contracted revenue showed up in the accounts — with the important difference that Micron owned its scarcity, and Nebius is renting someone else’s.
Frequently asked questions
What is Nebius’s possible price target?
The 18-analyst consensus is $226, which is about 19% below the $277.68 close on 14 August 2026 — the street currently thinks the stock has run ahead of itself. The street high is $410 from Northland Securities and the low is $120. The $356 figure in this article is not an analyst target: it is Micron’s current 12.2x trailing sales multiple applied to Nebius’s own guided $8bn year-end run-rate revenue, which implies roughly 28% upside.
Why does Nebius look so expensive on trailing numbers?
Because trailing numbers describe a company that no longer exists. Revenue grew 506.9% over the trailing twelve months to $1.36bn, so 56 times trailing sales is measuring today’s market value against a much smaller past business. Against the $3.0bn ARR Nebius had reached by June, the multiple is 25.4x, and against its $7bn–$9bn year-end run-rate guidance it is roughly 9.5x at the midpoint.
How is Nebius similar to Micron?
Both operate in genuine shortages where customers pay in advance to secure supply. Micron has 16 strategic customer agreements worth about $100bn in minimum contracted revenue through 2030. Nebius has customers prepaying 50%–60% of the capex needed to serve them. In both cases the customer is financing the supplier, which only happens when the scarcity is real rather than narrative.
How is Nebius different from Micron?
Fundamentally, and this is the main risk. Micron manufactures a physically scarce product and earns an 84.9% gross margin from a three-player oligopoly protected by manufacturing difficulty. Nebius rents GPUs it must buy from Nvidia, so its margins are structurally capped by its input cost. Micron self-funds capacity from $50bn of net income; Nebius is spending $20bn–$25bn a year against near-zero net income and must raise the difference.
What happens on 11 September 2026?
Nvidia’s contractual lock-up on its 9.3% Nebius stake expires. Nvidia holds 22,256,412 shares, mostly through a pre-funded warrant, and cannot exercise or sell before that date. Once it lifts, a large holder becomes mechanically free to sell. That does not mean it will — the passive 13G filing suggests otherwise — but the date is a known potential source of volatility.
What would make the bear case right?
Missing the year-end run-rate guidance is the main one: at $5bn rather than $8bn, Nebius would trade above Micron’s multiple and the valuation argument inverts. Beyond that, a falling prepayment ratio would signal the shortage easing, GPU supply catching up with demand would compress the $40m–$50m per megawatt short-term pricing, and the roughly $10bn a year of capex not covered by prepayments has to be financed in markets that may not always be open.
This article is for information only and is not investment advice. Prices, multiples and analyst targets are as of the close on 14 August 2026 and will have changed.