Predicting Microsoft and Alphabet earnings with AI: Compound vs the Street
Compound locked its Microsoft fiscal Q1 2027 and Alphabet Q3 2026 estimates before the release, then scored them four minutes after Microsoft reported
by Compound Team · October 28, 2026 · 7 min read
An earnings estimate is an analyst’s forecast of a company’s revenue and earnings per share for a quarter, and the consensus estimate is the average of those forecasts that a company beats or misses. On October 28, 2026, Compound predicted Microsoft’s and Alphabet’s earnings, emailed its locked estimates before the release, and was closer than the Street consensus on 3 of the 5 metrics that have one.
The challenge
Microsoft and Alphabet both reported after the close on October 28, 2026. At 12:19 pm ET that day, we gave Compound one instruction: predict both companies’ earnings with the best method it could find, and try to beat the Street. Lock the estimates and email them to the team before the release. As soon as the companies report, summarize the results, score the forecasts, and learn from the misses.
Here is how the day went, in ET:
- 12:19 pm. We asked for the forecasts.
- 12:34 pm. Compound locked its estimates and emailed them, with the forecast workbook, to the team.
- 4:15 pm. Microsoft’s 8-K posted. Compound monitors Microsoft, so the filing started the scoring work without anyone asking.
- 4:19 pm. Compound emailed the scorecard: closer than the Street on 3 of the 5 metrics that have a Street consensus. Microsoft was close to perfect; Alphabet was mixed. Compound saved what it learned for next quarter.
How Compound built the estimates
Compound started from its built-in market data: SEC filings, earnings call transcripts, consensus previews and Polymarket odds.
- Starting point. For Microsoft, its own guidance, restated in September for the new two-segment structure: revenue of $89.85-90.95B. Alphabet gives no guidance, so Compound built its revenue line by line from the last three quarters’ growth.
- Track record. Compound measured how far each company has landed from guidance and from consensus, then corrected for that pattern:
- Microsoft’s revenue beat its guidance midpoint by 1.5-3.2% in each of the last four quarters.
- Microsoft beat LSEG EPS consensus by 4-5%, leaving out last quarter’s one-offs.
- Alphabet beat LSEG revenue consensus by 2.2-2.5% in each of the last three quarters. Almost all of that came from Cloud, which beat its consensus by 9-11%.
- Alphabet’s EPS was in line or slightly below, because rising costs absorbed the extra revenue.
- This quarter’s news. Q3 PC shipments fell 22.4%, in line with Microsoft’s guide. Alphabet’s CFO flagged a slight FX headwind, said only a small part of TPU system revenue lands this year, and warned of margin pressure from renting outside capacity.
- The biggest call: Alphabet’s GAAP EPS would print well below what the market expected. Alphabet held $94.1B of SpaceX stock at June 30, and SpaceX shares fell from $170.86 to $150.86 over the quarter. Compound estimated a mark-to-market loss of roughly $11B, about -$0.69 per share after tax.
The locked forecasts
Compound locked these numbers at 12:34 pm ET. The 80% range is sized from how much the last three to four quarters’ surprises varied.
| Metric | Compound forecast | 80% range | Street |
|---|---|---|---|
| Microsoft revenue | $92.2B | $91.3-93.1B | $90.6B |
| Microsoft adjusted EPS (excl. OpenAI) | $4.90 | $4.76-5.05 | $4.71 |
| Microsoft Azure growth (constant currency) | 47% | 45.5-48.5% | 44-45% (company guide) |
| Alphabet revenue | $129.7B | $127.8-131.7B | $128.2B |
| Google Cloud revenue | $30.3B | $28.8-31.8B | $28.7B |
| Alphabet EPS excl. equity gains | $3.05 | $2.89-3.20 | $3.02 |
| Alphabet GAAP EPS | $2.35 | $1.90-2.80 | No consensus (Polymarket line: $3.10) |
Compound also took a side on four Polymarket markets at lock:
- Microsoft GAAP EPS above $4.82, priced at 81.5%: yes.
- Azure growth of 46-48% at constant currency, priced at 54%: yes.
- Google Cloud revenue above $30B, priced at 65%: yes.
- Alphabet GAAP EPS above $3.10, priced at 24.5%: no. Compound gave it a 6% chance.
The scorecard
The percentage in brackets is how far each forecast landed from the reported number.
| Metric | Reported | Compound | Street | Closer |
|---|---|---|---|---|
| Microsoft revenue | $92.28B | $92.21B (-0.1%) | $90.6B (-1.8%) | Compound |
| Microsoft adjusted EPS | $4.89 | $4.90 (+0.3%) | $4.71 (-3.7%) | Compound |
| Microsoft Azure growth (cc) | 48% | 47% | 44-45% (guide) | Compound (vs guide) |
| Alphabet revenue | $127.6B | $129.7B (+1.7%) | $128.2B (+0.5%) | Street |
| Google Cloud revenue | $30.7B | $30.3B (-1.3%) | $28.7B (-6.5%) | Compound |
| Alphabet EPS excl. equity loss | $2.67 | $3.05 (+14%) | $3.02 (+13%) | Street |
| Alphabet GAAP EPS | $2.12 | $2.35 (+11%) | $3.10 Polymarket line | n/a (market line) |
- Closer than the Street: 3 of the 5 metrics with a Street consensus (Microsoft revenue, Microsoft adjusted EPS, Google Cloud).
- Alphabet GAAP EPS, a prediction-market call: there was no Street consensus to score against. Compound gave GAAP EPS above $3.10 a 6% chance, against the market’s 24.5%, and it came in at $2.12. Compound’s call was right, and far more confident than the market’s.
- Inside the 80% range: 5 of 7 actuals. Alphabet’s revenue landed just under the range and its core EPS well under it.
- Polymarket calls: 3 of 4 right. Compound missed the Azure 46-48% bracket, because 48% resolves to the higher bracket.
Microsoft: close to perfect
Revenue was $92.28B against Compound’s $92.21B, and adjusted EPS was $4.89 against $4.90. The Street was below on both. Azure grew 48% in constant currency, one point above Compound’s 47% and above the company’s 44-45% guide. Operating income was $45.9B against Compound’s $45.7B.
Alphabet: right on Cloud and GAAP EPS, too high on revenue and core EPS
- Where Compound was right:
- Google Cloud revenue was $30.7B, up 103%. That beat the Street by $2B and landed 1.3% from Compound’s forecast.
- The SpaceX stake drove an equity-securities loss of $8.7B, or -$0.55 per share. Compound estimated -$11.0B. GAAP EPS came in at $2.12, far below Polymarket’s $3.10 line.
- Where Compound was wrong:
- Search grew only 12%, against Compound’s 15.6%.
- Unallocated AI costs were -$7.7B, against Compound’s -$5.4B.
- The core tax rate was about 19%, against Compound’s 14%.
- Operating income was $40.6B, against Compound’s $44.1B.
- Core EPS of $2.67 came in about 3 standard deviations below Compound’s distribution, so the Alphabet ranges were too narrow. Compound derives the $2.67 by adding back the $0.55 that Alphabet disclosed, so LSEG may score core EPS on a different basis.
What Compound learned
Compound saved these changes for the next earnings season:
- Weight the latest quarter’s beat more for guided KPIs like Azure.
- Treat a CFO’s “lapping an acceleration” warning as a real slowdown.
- Grow unallocated AI costs on their trend.
- Use a multi-quarter tax rate.
- Scale stock-price marks by about 0.8.
- Widen ranges for companies that give no guidance to at least ±10% on EPS.
- Call prediction-market brackets from the probability mass inside them, not from where the point estimate falls.
Run it yourself
Pick the companies you cover and have Compound forecast them before every print. For an analysis after every call too, set up a Watchlist.
-
Follow the company. In Company Universe, star the company and turn on Earnings and Filings in its event trigger. From the Compound CLI, the same step is one command:
compound agent watch MSFT --events earnings,filings -
Ask for a forecast. For example: “Predict MSFT’s quarter and beat the Street. Email me your locked estimates before the release, score them against the reported numbers as soon as they report, and keep what you learn for next quarter.”
Compound builds the model from filings, transcripts, consensus and prediction markets, emails the locked numbers, and wakes when the company reports to send the scorecard.
Sources: Microsoft and Alphabet Form 8-K earnings releases, October 28, 2026; Microsoft’s September 2026 8-K with restated guidance; LSEG consensus as reported in earnings previews; Polymarket odds at 12:34 pm ET. The Alphabet EPS excluding the equity loss is our calculation, not a company disclosure.
