Reference price: $463.47 (+9% in the September 2, 2026 premarket, after reporting results the night before). On a session like this any figure can go stale in minutes; numbers come from the earnings release filed with the regulator and from eight years of official accounts.
What this company actually does
It makes PCs and, above all, the racks of machines that go inside data centers. Two businesses: the traditional PCs for companies and consumers ($15.0B last quarter), and servers — where everything is happening: $31.8B, of which $16.4B is AI servers.
The thing to understand first: Dell does not make AI chips. It buys them, assembles them into a rack with cooling, cabling and storage, installs it and maintains it. It is the assembler, not the inventor.
The good — and something big happened last night
Revenue up 58%, operating profit up 3x. The summer quarter went from $29.8B to $47.0B (+58%) and operating profit from $1.8B to $5.4B (+204%). Out of every $100 of revenue, the business kept $6 a year ago; now it keeps $11.50.
And the reason dismantles the obvious objection. Everyone says AI servers carry a laughable margin — yet gross margin per $100 of revenue rose from $18.30 to $20.90. This quarter it kept MORE of every $100 than a year ago, not less. And the offices cost the same whether it sells $29.8B or $47.0B: operating expenses rose 21% while revenue rose 58%.
It raised full-year guidance by $25 billion overnight. Annual revenue: from $167B to $192B. AI servers: from $60B to $74B. EPS: from $17.31 to $24.37 (+181%). And it holds $95B of signed, unfilled orders — this quarter alone brought $60.9B of new AI-server orders, more than it billed in the entire quarter.
Two accounting-quality details: the gap between adjusted and GAAP earnings has shrunk from 60% to 5% in three years (old acquisition amortization is running out — the opposite of the dangerous case), and it has retired one in twenty shares in a year.
The bad
Problem 1 — and it is the finding of this analysis: profit multiplied by 3.5, and the real money didn't move. This quarter it reported $4.13B of profit (+255%) while $2.23B actually came through the door — less than last year. For the full half-year: reported profit ×3.5, free cash +0.2%. The fair part: over the full trailing twelve months free cash did grow 76% — it is not broken, but it grows at half the speed of reported profit and has stalled completely over the last six months.
Problem 2: half of that money is sitting in the warehouse. Inventory went from $10.4B to $21.3B in six months (+104%): from 36 to 52 days of sales. There is a good explanation — building an AI rack means buying the chips first, and they are extremely expensive; stockpiling is prudent with $95B of signed orders behind it. And a bad one: in any hardware business, inventory running ahead of sales is the first sign something isn't working out.
Problem 3: it is lending the money to its own customers. Customer financing receivables went from $14.3B to $20.4B (+43%). Translated: Dell sells the rack, books the entire sale as today's revenue, and lets the customer pay it off comfortably over years. That explains exactly why profit is up 255% while cash is flat: the profit has been counted; the money hasn't arrived.
Problem 4: it buys back shares with borrowed money. In the half-year it generated $4.1B of free cash and returned $6.8B in buybacks and dividends — $1.67 for every $1 generated, with $3.0B of net new debt. Not illegal, not unusual — but that "record capital returned to shareholders" wasn't paid by the business: it was paid by a loan.
Problem 5: on the books it is worth less than zero (equity of −$1.4B) — though the honest reading is different: it has spent $20.0B buying itself back, and the hole is closing. The real measure of its debt: $22.9B net against $14.3B of annual operating profit = 1.6 years, with interest covered about 20 times. Highly leveraged, but with the engine running.
Problem 6: the stock has already made the journey. From $114.43 at last fiscal year-end to $463.47: +305% in seven months. Earnings multiplied by 2.5 and the price by 4 — the difference is people paying more for every dollar it earns. This company traded its entire listed life between 4 and 13 years of earnings; today it costs 18. Add concentration ($60.9B of orders come from a handful of giants — if one cuts its plan, revenue doesn't dip, a chunk disappears) and an earnings yield of 2.83% while borrowed money costs 5–6%.
The one thing we hadn't seen in any other analysis this month
Usually, justifying what is expected of a company requires a margin it has never had. Not here. To earn the $28.07 expected for 2028, even with zero growth it would only need to repeat the margin it just achieved (on this year's revenue it would need to keep $11.18 of every $100 — it just did $11.50). And a second point in its favor: the panel says $23.16 for this year; the company guided $25.50 last night. Consensus is BELOW what the company itself promises — the multiples you see published are computed on earnings that are too low. The stock is somewhat cheaper than the screen suggests, not more expensive.
The right question
It is not whether the numbers add up. They do. It is this: Is the business really growing? Yes. Is the margin sustainable? Yes, if growth continues. Is what's being asked achievable? It is what it just did. Is the money coming in? For the last six months, no. Is it priced in? In good part.
Only one of two things can happen: either the inventory and the customer loans turn into cash over the next two quarters — and then this is a company that changed category at a still-reasonable price — or they don't, and what you are looking at is revenue booked in advance and financed with debt. Not both. And next quarter will say.
What we would do
Wait one quarter. And the reason is specific, not the usual generic caution: we are not waiting for an earnings number — the company has guided it and has never missed by much. We are waiting to see whether the money shows up: whether those $21.3B of inventory and $20.4B of customer loans convert into dollars. By November we will know.
| Price | What you would be paying for |
|---|---|
| Below $309 | Very attractive. |
| $309–393 | Attractive. |
| $393–505 ← it is here, at $463 | Fair value, upper end. |
| Above $505 | Expensive. |
With one honest caveat in its favor: if analysts raise their numbers ~10% to match last night's guidance — the normal course — that band shifts upward and $463 would sit in the middle rather than at the top edge. We are not saying it is expensive or that the business is bad: this is one of the few analyses this month where what the company is being asked to do is exactly what it just did. What holds us back is a 305% run in seven months, six months of cash not following profit, and a carrying cost of about three points a year while you wait.
The three things we would watch
1. The cash that comes through the door in November. This is THE number. It came in at $2.2B against $4.1B of profit. If next quarter cash approaches profit, the whole thesis is confirmed; if it stays at half again, you own a company that bills in advance and finances itself with debt.
2. Inventory and customer receivables. $21.3B parked and $20.4B lent out. Two more quarters growing faster than sales and it stops being preparation and starts being a collection problem.
3. New AI orders. $60.9B this quarter. It is the variable that rules everything and it doesn't depend on Dell: it depends on four or five companies continuing to sign construction budgets.
A note for readers who follow The Governor: the $20.4B Dell has lent its customers to buy AI servers is, on the other side of the contract, debt held by people building data centers — and Dell finances it by borrowing itself. This is not a warning about Dell: it is the AI-capex circuit seen from inside the supplier's balance sheet.
In one sentence
It assembles the racks where artificial intelligence lives and finally makes real money doing it — but it has doubled its inventory, lends its own customers the money to buy them, and buys back shares with what it borrows from the bank, so profit has multiplied by three and a half while cash hasn't moved in six months.