Have you ever watched a company print cash at a historic pace and still look, on paper, cheaper than it did when it was far smaller? That is the odd place Nvidia sits right now. The stock has climbed this year. The business has grown even faster. And the board just authorized another enormous pile of money to buy its own shares. In my view, that combination is the real story, not the headline number alone.
Why A Record Buyback Arrived With A Compressed Multiple
On Monday the company said it would add another $150 billion to its repurchase authorization. That sits on top of an $80 billion plan announced in May, when the quarterly dividend also jumped from a token penny to 25 cents a share. Those are not symbolic gestures. They are a management team telling the market, in the only language markets reliably hear, that the stock looks inexpensive relative to the cash the business is about to generate.
Look at the forward picture. The price-to-earnings ratio tied to fiscal 2028, which starts next February, sits around 14.5. That is below most other megacap names and well below the firm’s own five-year average current multiple of about 62.9. I’ve found that investors often treat a falling multiple as a warning. Sometimes it is. Sometimes it is just earnings outrunning the share price.
Buying back the stock is a tremendous opportunity.
– Company leadership, speaking publicly last month
That line was not a throwaway. The same executive later described the firm as the world’s first growth value stock. People keep trying to shove it into one box or the other. Growth. Or value. The argument now is that it can be both, because expected profits are scaling so quickly that a mid-teens multiple still prices in a lot of skepticism.
The Earnings Math Behind The Bargain Claim
Analysts, on average, see net income approaching $385 billion in fiscal 2028. That would be roughly 60 percent higher than the prior year and more than five times the level from three years earlier. Sales growth guidance for that same fiscal year has been sketched near 70 percent. Let that sit for a second. A company already valued above $5.5 trillion is still talking about that kind of top-line expansion.
The stock is up about 23 percent this year and has beaten the Nasdaq. Fine. It has not kept pace with the earnings path those numbers imply. That gap is why the multiple looks tight. It is also why a repurchase program of this size is not just financial engineering. It is an allocation decision: management would rather own more of its own cash machine than park the money elsewhere.
One portfolio manager put it plainly. If earnings are scaling faster than the share price, buying the stock can be the best internal investment available in the coming year. I tend to agree, with one caveat. Buybacks only create lasting value if the shares are truly cheap and the cash would not earn more inside the product roadmap. On current forecasts, that bar looks reachable.
How The Authorization Could Shrink The Share Count
The company had already said it wanted to return about half of free cash flow through repurchases and dividends. Spend the full current authorization and the share count could fall by roughly 4 percent. That is not a rounding error on a name this large. Every point of shrinkage lifts per-share earnings if profits hold.
Research notes circulating after the announcement suggested the stepped-up program could add several cents to calendar 2027 earnings per share. The exact figure will depend on timing, average purchase price, and how quickly cash actually hits the register. Still, the direction is obvious. Fewer shares. Same profit pool. Higher EPS.
- Additional $150 billion authorized on top of the May plan
- Dividend raised earlier to 25 cents from 1 cent
- Target of returning about half of free cash flow
- Potential mid-single-digit reduction in shares outstanding
Monday’s tape reaction was modest, about a 2 percent bounce, alongside news of new software and hardware tools meant to control AI agents. The market heard the buyback. It did not immediately re-rate the whole story. That lag is part of why the multiple remains compressed.
A Multiple That Sits Below Almost Every Peer
Compare the fiscal 2028-style multiple with other giants measured over a similar window. Apple sits near 35.5. Alphabet near 22.6. Microsoft near 21.7. Amazon near 23.2. Among chip names tied to data centers, Broadcom is around 18.2, AMD around 38.2, and Intel around 54.7. Nvidia’s 14.5 looks lonely at the low end.
None of those rivals is forecasting 70 percent sales growth next year. That matters. Broadcom’s silicon story leans on custom designs for a handful of huge customers. AMD competes in GPUs but with a sliver of the installed base. Intel still lives mostly in central processors and has not fielded a GPU lineup that truly matches the current training and inference stack. Different businesses. Different growth. Similar or richer multiples.
| Company lens | Forward-style P/E | Near-term growth backdrop |
| Nvidia | About 14.5 on FY28 | Very high sales and profit growth guided |
| Other megacaps | Low 20s to mid 30s | Solid, generally slower |
| Chip peers | High teens to 50s | Mixed GPU and custom-silicon stories |
Is the comparison perfectly clean? No. Accounting years differ. Mix of businesses differs. Risk of a demand air pocket differs. Even so, the gap is wide enough that calling the stock “expensive because it is Nvidia” no longer does much work. The market is already applying a discount for the fear that growth must slow.
Why Investors Still Struggle With The Slope Of Growth
Here is the uncomfortable part. After a monstrous run, people find it hard to believe the slope can stay this steep. One investor framed it as discomfort with the downward curve of the growth rate, even if the absolute dollars keep exploding. That fear is rational. Nothing compounds at 70 percent forever. The question is whether “slower” still means “exceptional,” and whether 14 times those later earnings already bakes in a hard landing.
I’ve sat with enough earnings seasons to know that markets hate deceleration more than they love huge absolute profits. A company can print record cash and still de-rate if the second derivative turns. That is the psychological tax Nvidia is paying. The buyback is, in part, an attempt to argue with that tax using cash instead of slogans.
It is just really hard for investors to get comfortable that this pace will continue. That downward slope in the growth rate is why the multiple looks so tight.
Fair. Comfort is not a valuation input, though. Cash flow is. If fiscal 2028 lands anywhere near the consensus profit figure, today’s price is not pricing a victory lap. It is pricing a debate.
Huang’s Public Case For Owning More Of The Company
Leadership has been unusually direct. The stock is “misunderstood,” the argument goes, because the market cannot decide whether to pay for growth or for value. The preferred answer is both. High growth now. High cash later. A multiple that still looks like a mature industrial even while the product cycle is mid-boom.
On a morning show the same week as the authorization, the message was simple. The firm will generate a lot of cash in the coming years. As that cash arrives, more of it should go back to owners. Not every dollar. Product and ecosystem spending still come first. But the residual is no longer an afterthought.
Perhaps the most interesting aspect is how blunt the “bargain” language has become. CEOs usually hide behind “we like our shares at these levels.” This one has said, in effect, that repurchase is one of the best uses of capital on the menu. That is a tell. Boards do not authorize $150 billion extra because they are bored.
What “Growth Value” Actually Means In Practice
The phrase sounds like marketing. It is not empty if you unpack it. Growth value, in this usage, means a business still expanding at a rate associated with young platforms, while already throwing off the cash associated with mature platforms. Rare combination. When it appears, multiples often stay confused for a long time because models are built for one regime or the other.
- Growth investors want another re-rating on the sales line.
- Value investors want a floor based on owner earnings and buybacks.
- The stock currently satisfies neither camp completely, which is why it can look cheap to both.
In my experience, those in-between years are when patient capital does the most work. You do not need a new narrative. You need the existing narrative to keep converting into cash. Repurchases then act like a slow, automatic concentrator of that cash into fewer hands.
Cash Flow, Not Hype, Is The Constraint Now
Demand for the graphics processors that train and run large models remains the engine. That demand has already rewritten the income statement. The next chapter is less about whether the chips sell and more about how cleanly the cash converts after inventory, customer concentration, and the usual supply-chain noise.
Guidance through early 2028 already implied hundreds of billions more in sales than many models had assumed earlier in the cycle. If that path holds, the repurchase math gets easier every quarter. If it slips, the authorization simply lasts longer. Either way, the board has pre-committed to treating the equity as an asset worth accumulating.
Does that guarantee a higher multiple? No. Multiples are moods. Earnings are arithmetic. Buybacks operate on the arithmetic. Over a long enough stretch, arithmetic tends to win arguments that slogans cannot.
How A Bigger Buyback Can Nudge Valuation Higher
One analyst with a buy stance argued the stock “deserves to be higher given its growth rate,” and that larger, repeated repurchases can help close that gap. I think that is half right. Buybacks do not create demand for chips. They can, however, keep per-share results rising even if the multiple stays stubborn. Eventually some allocators notice the EPS and the cash yield and stop treating the name like a story stock.
There is a second, quieter effect. A shrinking float makes each remaining share a slightly larger claim on a business that already dominates training clusters. That is not magic. It is ownership math. Combined with a dividend that is no longer decorative, the capital-return package starts to look like something a quality compounder would run, not a speculative flyer.
Simple owner-earnings sketch: Exploding operating cash Minus maintenance and growth capex Equals a surplus large enough to fund both product bets and a historic repurchase book
That sketch is why the “too cheap to resist” framing is not just color. If the surplus is real, buying the shares is internally consistent. If the surplus is a mirage, the authorization becomes a very expensive signal. Current public forecasts still point to the first case.
Risks That Could Make The Cheap Multiple Look Fair
Let’s not pretend the discount is only confusion. Customers could build more of their own silicon. Utilization of installed GPUs could disappoint. Export rules could bite. Pricing could ease once supply catches demand. Any of those would flatten the earnings path that makes 14.5 times look silly.
There is also the simple problem of scale. At more than $5.5 trillion, even excellent execution moves the needle less than it did three years ago. A 70 percent sales jump from a huge base is a different animal than the same percentage from a smaller one. Execution risk rises with the number of zeros.
I would not ignore customer concentration either. A handful of cloud and model builders account for an outsized slice of data-center demand. Their capex cycles will not stay synchronized forever. When one pauses, the tape will treat it as a thesis break even if the multi-year buildout is intact.
- Custom silicon at large buyers
- Slower GPU utilization after the first wave of builds
- Policy and export friction
- Normal pricing gravity once scarcity fades
- The raw difficulty of growing off a multi-trillion base
Those are real. They are also, to a degree, already visible in the multiple. You do not get a mid-teens forward earnings claim on a dominant franchise unless the market is nervous. Nervous can be right. Nervous can also be early.
What The Dividend Change Quietly Signals
Raising the quarterly payout from 1 cent to 25 cents does not turn this into an income vehicle. It does mark a cultural shift. Token dividends say “we are still a builder.” A meaningful dividend plus a record repurchase book says “we are a builder that now admits the cash is too large to reinvest at the same hit rate.”
That admission is healthy. Companies that refuse to return surplus often end up forcing it into lower-return projects. Companies that return too much can starve the roadmap. The stated mix, about half of free cash flow, tries to live in the middle. Whether that split stays sacred will tell you more than any single authorization headline.
Reading Monday’s Tape Without Overfitting One Session
A 2 percent pop is not a re-rating. It is a nod. Markets had already heard the growth story in August. They had already heard the “misunderstood” line at a conference earlier in the month. The new information was size: another $150 billion of permission to buy paper the board thinks is inexpensive.
Software and hardware for managing AI agents landed in the same release cycle. That product news matters for the long ecosystem argument. It did not, by itself, change the 2028 earnings power. The buyback did not change 2028 earnings power either, except at the per-share level. Together they said: the platform is still expanding, and the residual cash has a home.
If you needed a single-session verdict, you will be disappointed. The interesting work happens across the next several quarters of actual repurchase activity. Authorizations are promises. Open-market buying is behavior.
How I Frame The Position Size Question
This is not a recommendation to swing at the name tomorrow morning. It is a framework. If you already believe the data-center build lasts into 2028, the compressed multiple plus an aggressive repurchase book improves the expected value of holding. If you think the build peaks much sooner, a cheap multiple on fading growth is a trap, and the buyback only delays the recognition.
Position size, then, should track your confidence in duration, not your affection for the brand. Duration is the whole game. Three more years of outsized cash conversion makes today’s price look careless. One more year of outsized cash followed by a stall makes it look merely fair.
Buying the stock in a bigger and bigger way can help solve the valuation problem created by a growth rate the market refuses to pay for in full.
That is the cleanest version of the bull case I can write without cheerleading. Help. Not guarantee. Solve. Not erase. Markets can stay skeptical longer than a repurchase program can run. They usually do not stay skeptical forever if the cash shows up on schedule.
A Practical Checklist Before You Treat The Dip As A Gift
Before anyone treats every pullback as an engraved invitation, run a short list. Not a model. A list. You can keep it on a note card.
- Are large customers still lifting capex guidance, or only defending last quarter’s number?
- Is gross margin holding as supply improves, or is price the hidden variable?
- Is the repurchase actually printing in the cash-flow statement, not just sitting in a press release?
- Is the share count falling after dilution from employee stock, or only before it?
- Does fiscal 2028 commentary still support the profit figure that makes 14.5 times look cheap?
If those answers stay constructive, the authorization is a tailwind. If they wobble, it becomes a cushion, not a catalyst. Cushions are useful. They are not the same as upside.
The Decade-Low Multiple In Context
Calling the stock the cheapest in a decade on one metric is catnip for headlines. Context still matters. Ten years ago the business was a different animal. Gaming and visualization carried more of the identity. The data-center mix was smaller. Comparing today’s forward multiple with yesterday’s trailing multiple is a sloppy habit, even when the direction of the comparison is honest.
What you can say cleanly is this. Relative to its own recent history, and relative to peers that are growing more slowly, the claim on 2028 earnings is tight. Relative to the risk that 2028 looks nothing like the current sketch, the claim may be appropriate. Both sentences can be true at once. Adults can live with that.
I’ve found that the market pays up again only after two or three reports in a row that refuse to break. One beat is noise. A pattern is a repricing. The buyback does not replace that pattern. It sweetens the wait.
Capital Return As A Confidence Signal, Not A Substitute For Demand
Some readers will call the authorization a stunt. I do not. Stunts are small. This is large enough to change per-share math if executed. At the same time, no repurchase program invents a new customer. The GPUs still have to ship. The software layer still has to matter. Agents still have to become a workload that needs control planes and accelerators.
Think of the buyback as a second engine, not the first. The first engine is demand for compute. The second is the decision to let owners keep more of each dollar that demand creates. Two engines beat one. One engine with a broken second is still an airplane. Just a slower one.
That analogy is imperfect, I know. Fine. The point stands. Do not let a record authorization distract you from unit demand, and do not let unit demand make you ignore a board that is willing to retire equity at what it believes is a discount.
Putting The Whole Picture On One Page
So where does that leave a reader who has to decide whether this is a curiosity or a setup? The company is authorizing historic repurchases while its forward earnings claim sits well below its own recent average and below most megacap and chip peers. Leadership is saying out loud that the shares look like a strong use of cash. Consensus profits for fiscal 2028 are enormous. The market is still allergic to the idea that growth can stay this rich off this base.
That is the entire tension. Not a mystery. A tension. If you need the multiple to expand next month, you may wait. If you can live with ownership of a cash machine that is buying itself on the cheap while the argument drags on, the structure is clearer than it has been in a while.
I keep coming back to a simple line. Earnings are scaling faster than the share price. When that happens at a dominant franchise, buybacks stop being a footnote. They become part of the thesis. They do not complete the thesis. Demand still has to do the heavy lifting. But they change the slope of the per-share path, and that is often how long debates finally get settled.
Watch the cash. Watch the share count. Watch whether 2028 commentary still supports the profit figure that makes today’s multiple look severe. Everything else is noise dressed up as narrative. And narrative, unlike a repurchase print, does not retire stock.
