Earnings estimates are not just forecasts — they are a live signal of how the market is repricing the future. Watching how Wall Street revises those estimates over time can tell you more about a company's trajectory than almost any other single metric. A close look at Nvidia and Apple illustrates this principle sharply, and the gap between the two is more dramatic than most investors appreciate.
Why Earnings Estimate Revisions Matter
The core idea is straightforward: when analysts systematically revise earnings estimates upward, it means the business is outperforming their models. That revision process — not just the current price or P/E ratio — is often where the real signal lives. A fund philosophy worth taking seriously holds that earnings estimate revisions are the single most important variable in stock selection. Whether or not you take that claim at face value, the direction and magnitude of revisions are undeniably important data points.
The practical exercise is to look at long-dated estimates — 2025 through 2030 — and ask how much they have moved, and in which direction.
Nvidia: Estimates That Keep Getting Beaten
Nvidia's estimate revisions tell a remarkable story.
38:45
Spreadsheet showing Nvidia earnings estimates by calendar year with recent revisions highlighted
Watch at 38:45 →
- Calendar year 2025: Estimates moved from $4.50 to $4.62 — roughly a 3% bump.
- Calendar year 2026: The street moved from $6.42 to $7.50. That is a nearly 17% upward revision — and the new number is likely still too low.
- Calendar year 2027: Estimates rose by almost $2 per share, a roughly 30% increase. At $9.65, the stock trades at only 21 times 2027 earnings.
- Calendar year 2028: The revision here is staggering — estimates moved from approximately $8.59 to $11.00, nearly a 50% miss by Wall Street corrected in real time.
- Calendar year 2030: Early estimates sit around $13, implying roughly 15 times 2030 earnings.
Running those numbers through a CAGR calculation, Nvidia's earnings are growing at approximately 23% annually through 2030 — and that estimate is probably still too conservative. The street has consistently underestimated Nvidia's earnings power at every horizon. There is a strong case that the 2026 estimate will effectively be beaten in 2025 fiscal reporting.
Here is the critical valuation insight: the broader S&P 500 trades at roughly 24 times earnings and is growing significantly slower. Strip Nvidia out of the index and the average growth rate drops further. Yet Nvidia trades at 21 times 2027 earnings. The market is paying more for slower-growing companies than it is for Nvidia on a forward basis.
Apple: Solid but Slower
Apple's estimate revisions tell a different story.
46:10
Spreadsheet comparing Apple earnings estimates across calendar years 2025 through 2030
Watch at 46:10 →
- Calendar year 2025 estimates moved up modestly — about $0.20.
- Revisions for 2026 and 2027 were similarly incremental.
- The 2029 estimate saw a more meaningful revision — roughly a dollar, or about a 10% bump.
Apple's earnings CAGR through the forecast period comes in around 10% — solid for a mature mega-cap, but less than half of Nvidia's 23%. Apple is a generational business, but the revision cadence does not suggest the same kind of systematic underestimation that characterizes Nvidia's history.
SAP: The Enterprise Software Survivor
SAP is a useful comparison case for thinking about software durability. The company went from roughly €22 billion in revenue in 2015 to an expected €36–37 billion in 2025 — steady but unspectacular growth for most of that period. The critical question for SAP was whether it would be disrupted by the SaaS transition. It wasn't. It became SaaS itself and is now growing cloud revenue at 27% year-over-year.
At approximately 5 times revenue and 26 times earnings, SAP is not obviously cheap, but it is arguably more defensible than it looks. Analysts note it is among the least exposed enterprise software companies to generative AI cannibalization, largely because ERP implementation is so deeply customized and complex that replacing it — even with AI-assisted coding — is not a realistic near-term threat for large enterprises.
28:30
SAP revenue and valuation metrics in local currency (euros) showing enterprise value to sales ratio
Watch at 28:30 →
The German legal context adds color here: in Germany, a Series A or private financing agreement requires the entire contract to be read aloud before a notary — a process that can run six hours. The cultural seriousness around contractual commitment embedded in SAP's home market arguably reflects why ERP software, once installed and customized over years, is almost never simply replaced.
How to Think About Long-Dated Estimates as an Investor
A few practical takeaways from this kind of analysis:
- Watch the direction of revisions, not just the level. A stock with a high P/E but consistently rising estimates may be cheaper than it looks on a forward basis.
- Long-dated estimates are often wrong — and that's useful. When the street is systematically too low at the 2–4 year horizon, it suggests the market has not yet priced in the full growth runway.
- The market multiple is a benchmark, not a ceiling. If the S&P trades at 24 times earnings with low single-digit growth and a hyper-growth company trades at 21 times two-year-forward earnings, the relative valuation argument is straightforward.
- Patents are an underrated learning resource. For anyone trying to understand a new industry — semiconductors, biotech, photonics — reading patents provides structured, required-to-teach technical disclosure that most investors overlook entirely.
The broader point is that earnings estimates, tracked over time and across vintages, function almost like a running audit of how well a business is executing against expectations. Nvidia has failed that audit in the best possible way, repeatedly and dramatically. That pattern is rarely random.
