Should Meta keep funding Reality Labs?
Recommendation. Keep funding Reality Labs only as a smaller, gated bet on AI glasses: stop VR and Horizon in 2027, cut AI-glasses spending to about $6.9bn a year, half of this year's level, for 2027 and 2028, and carry on after the FY2028 10-K only if AI-glasses revenue has grown at least 55% a year. At its current size Reality Labs does not pay back in any of my three scenarios: against shutting it down, keeping it as planned is worth −$83.3bn in the base case and still −$59.0bn if AI glasses become a mass product. The gated plan loses at most $8.9bn if AI glasses disappoint and is worth $51.4bn if they take off, so it beats stopping whenever that bull case has more than a 15% chance (with bear and base otherwise equally likely).
What the money has bought so far. From 2019 to 30 September 2026 Reality Labs reported $100.9bn of operating losses and $13.6bn of revenue. Revenue was $2.27bn in 2021 and $2.21bn in 2025: four more years and $66.8bn of further losses left it 2.9% lower. The FY2021 and FY2022 10-Ks said Reality Labs investment would increase, the FY2023 and FY2024 10-Ks said its operating losses would too, and the loss rose every year, from $10.19bn in 2021 to $19.19bn in 2025. The FY2025 10-K said 2026 would be similar, and the first nine months are 3.0% below last year's. No filing states a revenue goal; since 2023 each one says Reality Labs has "generated only limited revenue." Counting the spending through the end of 2026, with this year's fourth quarter estimated, and carrying it at 10% after tax, the money spent would have to earn $11.0bn a year just to cover its cost of capital; the last four quarters lost $14.9bn after tax. That money is sunk and is not part of the decision below, but it is the record the next budget should be held to.
The four options. Each option is valued against shutting Reality Labs down in 2027, after tax, at a 10% discount rate, from 2027 to 2036 plus a terminal value.
| NPV against stopping | Bear | Base | Bull |
|---|---|---|---|
| Keep funding as planned | −$86.7bn | −$83.3bn | −$59.0bn |
| AI glasses only, full budget | −$62.5bn | −$59.0bn | −$17.3bn |
| AI glasses only, half budget, gate in 2028 | −$8.9bn | −$8.7bn | $51.4bn |
| Stop Reality Labs | $0.0bn | $0.0bn | $0.0bn |
Keeping Reality Labs as planned never pays back within ten years in any scenario; to break even in the base case, AI-glasses revenue would have to grow about 93% a year for five years. Funding AI glasses alone at the full budget is better but still loses money in the bull case, because it carries about $13.7bn a year of fixed spend; it would need about 85% growth a year. Stopping costs about $5.2bn once, before tax, and nothing after that. The gated plan loses $8.9bn in the bear case and $8.7bn in the base case, because the 2028 review shuts it in both; in the bull case it returns 36% a year and pays back in 2036.
The options I rejected. I rejected keeping Reality Labs as planned, which is Meta's stated plan for 2026, because no scenario I can support from the filings earns back its cost. Spend growth and the discount rate move its value most, and even the best pairing of the two that I tested, spend falling 5% a year and a 12% discount rate, since a higher rate shrinks today's value of future losses, leaves it at −$59.4bn. Stopping everything now is the closer call, and it is the right answer if you give AI glasses less than a 15% chance of becoming a mass product. I chose the gated plan over it for three reasons. Its worst case, $8.9bn, is small next to the $83.3bn the current plan costs in the base case. AI glasses are the only part of Reality Labs whose sales are growing: six of the seven filings since Q1 2025 say the change in revenue came from AI glasses up and Quest down. And Meta is already moving this way, with 70% of its 2026 Reality Labs spending going to wearables and Reality Labs employee pay in Q3 2026 24% below a year earlier.
Stop conditions for the gated plan.
- Every quarter: stop if Reality Labs' operating loss, leaving out wind-down charges, is above $1.80bn two quarters running: 10% over a quarter of the plan's base-case 2027 loss of $6.55bn.
- On the FY2027 10-K: stop a year early if 2027 AI-glasses revenue is below $1.74bn.
- On the FY2028 10-K: stop unless 2028 AI-glasses revenue is at least $2.71bn, growth of 55% a year from an estimated $1.12bn in 2026, at a contribution margin of at least 30%. With a lower margin the growth needed is higher; the gate map on the Gate sheet shows the trade.
- Review at once if Meta books another loss on non-cancelable Reality Labs inventory commitments, as it did in 2025: it is the first sign that demand is below plan.
Once VR stops, Reality Labs revenue in the 10-Q is almost all AI glasses, so the second and third tests can be read from outside the company; inside Meta they can be read every month.
What would change my view. Three things would make me fund AI glasses more, not less: 2027 glasses revenue above the bull path's $1.91bn; evidence that the margin is above 30%, for example if Meta began reporting Reality Labs cost of revenue; or proof that the glasses bring in money the segment is not credited with, such as use of Meta AI that lifts Family of Apps revenue. For funding AI glasses at the full budget to break even in the base case, that outside benefit would have to be about $9.6bn a year after tax. A partner paying part of the development would work like a smaller budget.
Risks and what the model leaves out. The model values only Reality Labs' own revenue. If AI glasses turn out to be how people reach Meta's AI, losing that position could cost the rest of Meta more than anything counted here; that is the case for keeping the gate open instead of stopping. Halving the budget may also slow the product. The model does not let a smaller budget slow growth, so it flatters the cut, and the 2028 gate is what would catch it. Meta does not disclose glasses revenue, margin or cost, so the 50/50 revenue split between AI glasses and Quest, the margins and the growth rates are my judgments; the 70/30 cost split is Meta's statement for 2026 operating expenses, applied here to all costs. The bull case is large: its 2036 AI-glasses revenue of $57.1bn is 1.6 times Apple's fiscal 2025 Wearables, Home and Accessories sales ($35.7bn), and all of its value comes from after 2036: the terminal value alone is worth $54.0bn. Losses are valued after a 21% tax saving, which assumes the rest of Meta stays profitable, and at a 10% discount rate against effective rates of 4.27% to 5.77% on Meta's newest notes.
Basis. Every number comes from Meta_Reality_Labs_Investment_Model.xlsx after Excel recalculation (all 21 checks OK). The facts are from Meta's 10-K filings for 2021 to 2025 and its 10-Q filings through Q3 2026 (latest accession 0001628280-26-066773, filed 22 October 2026); each accession number is on the Inputs and History sheets. 2026 is estimated as the first nine months plus last year's fourth quarter grown at the nine-month rate, which gives an operating loss of $18.6bn. The three scenarios differ in four judgments, set out on the Inputs sheet: AI-glasses growth for 2027 to 2031 (10%, 35%, 70% a year), Quest's yearly change, long-run growth, and the contribution margin (10%, 20%, 30%). This is an exercise on public filings, not investment advice.
