Meta vs Tesla Q3 2026: $70B+ AI Capex Splits Wall St

US stock futures slid on Thursday, October 22, 2026, after Meta and Tesla reported third-quarter results that pointed in opposite directions on AI spending. Meta beat on revenue and earnings, then narrowed its 2026 capital-expenditure range to $135 billion to $145 billion. Tesla beat on adjusted earnings, then said it would spend about $20 billion this year. For developers, the split matters more than the headline index moves, because the money these two companies commit decides which coding tools get cheaper, which get rationed and which stay experiments.

This analysis covers the market reaction, the numbers from both reports, and what the Meta Tesla earnings contrast means for the AI coding tools that engineers use every day. It also sets out five predictions you can check against the next two quarters.

Google · Preferred Sources

Don't miss new tech stories on Google

Add TrendinTech once in the Google app and our stories appear in your news suggestions.

Add Now

Futures Slip as Meta and Tesla Report: What Happened Overnight

Both companies reported after the close on Wednesday, October 21. By early Thursday, S&P 500 futures were down 0.5% in CNBC’s market coverage, while another report put the decline at 0.4%, to 7,768.25 points. Nasdaq-100 futures fell 0.7% to 29,800.00 points. Dow Jones Industrial Average futures dropped 0.3%, or 170 points.

The tone was set by the AI trade rather than by a single miss. Swissinfo’s markets wrap described a tech earnings rally that faded overnight, and The Financial Express framed the session around AI concerns and the two earnings reports. Earlier in the week the mood had been better. Investopedia’s Wednesday wrap recorded major indexes rising as tech stocks surged, with SpaceX jumping after its own results.

Tesla shares traded lower after hours. CNBC’s account had them down 2%, while Bloomberg-syndicated coverage put the drop at 3.3%. The two figures were captured at different moments in thin trading, so treat the exact decline as unsettled. Meta’s after-hours move is not part of the figures we can confirm, which is itself a signal. The story traders focused on was Meta’s bill and Tesla’s retreat, not a clean beat or miss.

Index futures (Oct. 22, 2026)	Reported move	Level or size
S&P 500 futures (CNBC)	Down 0.5%	Not stated
S&P 500 futures (other report)	Down 0.4%	7,768.25 points
Nasdaq-100 futures	Down 0.7%	29,800.00 points
Dow Jones futures	Down 0.3%	170 points
Tesla after hours (CNBC)	Down 2%	Share price not stated
Tesla after hours (Bloomberg-syndicated)	Down 3.3%	Share price not stated

Meta reported third-quarter revenue of $64.60 billion, up 26% from a year earlier. The LSEG consensus estimate was $63.37 billion. Earnings came in at $3.83 per share against an analyst estimate of $3.57. In Meta’s own release, family daily active people averaged 3.64 billion in September, up 3%. Ad impressions rose 10% and the average price per ad climbed 14%, which means pricing did more of the work than volume.

Meta guided fourth-quarter revenue to $73 billion to $76 billion. Against the $64.60 billion just booked, that implies sequential growth of roughly 13% to 18%. The company said it expects 2026 operating income to exceed 2025 operating income if you exclude the legal charge discussed below.

The beat was real, but it came with a bill. Meta’s total costs and expenses rose 68% to $51.61 billion, and headcount fell 15% year over year to 66,775 as of September 30. The company is shrinking its workforce while it grows its compute budget. That trade, fewer people and more GPUs, is the pattern engineers should watch because it shapes hiring, tooling budgets and the expectation that each remaining developer ships more with agents. Our report on developers who let AI write half their code covers how that expectation is landing on the ground.

The $9.89 Billion Charge Behind Meta’s Headline EPS

Meta’s third-quarter general and administrative expenses include a $9.89 billion charge tied to an August 2026 agreement with US state attorneys general. That agreement resolved litigation over teen mental health and social media. Income from operations was $12.99 billion, and Meta’s release says it would have been $22.88 billion without the charge. Diluted EPS of $3.83 would have been $7.40 on the same basis.

This matters for reading the beat. Analysts’ $3.57 estimate and Meta’s $3.83 result both sit below the adjusted figure, so the comparison tells you the charge was largely anticipated. The underlying operating engine looks strong. The cost side looks heavier than a year ago even after you strip the charge, because the AI build-out is now a recurring expense rather than a one-time project. We covered the earlier legal fallout in our piece on the $25 billion New Mexico privacy verdict, and the Q3 charge shows that legal costs and AI costs are now competing for the same cash.

Meta’s AI Capex: $135 Billion to $145 Billion and Negative Free Cash Flow

Meta now expects 2026 capital expenditures, including principal payments on finance leases, of $135 billion to $145 billion. That narrows the prior outlook of $130 billion to $145 billion by lifting the floor. Third-quarter capex alone was $34.91 billion. Cash flow from operating activities was $32.79 billion, so free cash flow came out at negative $2.12 billion. Cash, equivalents and marketable securities stood at $82.45 billion on September 30.

Meta also raised its full-year total expense outlook to $175 billion to $179 billion to absorb the legal charge. It said Reality Labs losses should decline to a range of $18 billion to $19 billion this year and narrow to around $15 billion next year as it reallocates investment toward other priority areas, primarily AI. In plain terms, the virtual-reality budget is being drained to fund model training and inference.

The simple arithmetic below shows how different the two companies’ spending intensity is this quarter. It uses only the reported figures.

# Q3 2026, figures in $ billions as reported
meta_rev, meta_capex, meta_ocf = 64.60, 34.91, 32.79
tsla_rev, tsla_capex = 28.05, 4.73

print(round(meta_capex / meta_rev * 100, 1))   # 54.0 -> % of revenue spent on capex
print(round(tsla_capex / tsla_rev * 100, 1))   # 16.9
print(round(meta_ocf - meta_capex, 2))         # -2.12 -> Meta free cash flow

Meta spent about 54% of quarterly revenue on capex. Tesla spent about 17%. Both numbers are large, but only one company is paying for its build-out by drawing on its balance sheet in the quarter. Our earlier Meta Q3 earnings preview flagged the $41.68 billion AI spend question that this report has now partly answered.

Tesla’s Q3: $28.05 Billion in Revenue and a Capex Cut to $20 Billion

Tesla reported third-quarter revenue of $28.05 billion and adjusted earnings of $0.48 per share. CNBC listed the analyst revenue estimate as $27.88 billion, while other coverage cited $27.92 billion, so the beat is slim either way. Teslarati’s tracker put the adjusted EPS estimate at $0.45. Operating income was $718 million on a 2.6% operating margin, and free cash flow was $1.531 billion. Secondary coverage from Electrek and Investing.com described operating income as down 56% from last year.

The capex story is the real news. Tesla said capital expenditures would be approximately $20 billion for the year, and one report said earlier guidance had been above $25 billion. Third-quarter capital spending was about $4.73 billion, against roughly $2.25 billion a year earlier, and analysts had expected a figure close to $8 billion. Nine-month capex was about $13.01 billion, so a $20 billion year implies roughly $7 billion in the fourth quarter.

Research and development tells the opposite story. R&D was $2.76 billion, up from $1.63 billion in Q3 2025, a rise of 69%. Tesla is spending less on concrete and more on engineers and training runs. We analyzed the cash-flow side in depth when we wrote that Tesla’s $1.53 billion cash flow arrived as AI capex fell short, and the revenue mix in our look at Tesla’s $28 billion quarter and FSD software growth.

Meta vs Tesla: The Q3 Scoreboard

Metric (Q3 2026)	Meta Platforms	Tesla
Revenue	$64.60 billion (up 26%)	$28.05 billion
Revenue vs estimate	$63.37 billion (LSEG)	$27.88 billion (CNBC) or $27.92 billion
Earnings per share	$3.83 vs $3.57 expected	$0.48 adjusted vs $0.45 expected
Q3 capital expenditure	$34.91 billion	About $4.73 billion
Full-year 2026 capex outlook	$135 billion to $145 billion (floor raised from $130 billion)	About $20 billion
Free cash flow	Negative $2.12 billion	$1.531 billion
R&D or headcount signal	Headcount down 15% to 66,775	R&D $2.76 billion, up 69%

The table shows two companies making very different bets. Meta is the larger spender on physical infrastructure and the weaker generator of free cash this quarter. Tesla spent far less on capex than the Street feared and produced positive free cash flow, yet its operating margin remains thin at 2.6%.

Two Opposite AI Bets, Explained

Meta is betting that frontier models can be sold directly, to consumers and to developers. Tesla is betting that AI will live inside machines it already builds. Those theses produce different spending shapes. Meta needs data centers because every request to a model costs compute that it must own or lease. Tesla needs training capacity and factories, but its revenue today comes from cars and energy storage, not from selling inference.

Tesla’s Q3 materials reportedly described Super Intelligence training compute as a major infrastructure item, with a target of 400 MW by December 2026, up from about 230 MW in September. Reports also said Tesla expects to begin commissioning equipment at its own Austin chip plant in 2027. Those details come from secondary coverage, and Tesla has not published a priced AI product for developers. On the Optimus humanoid robot, Tesla said it had begun producing the first production-quality units at Fremont, and Elon Musk said customer deliveries could begin near the end of 2027. That is a target, not a launch date. Our look at the earlier Optimus production claims shows how far those goals have stretched before.

Tesla and SpaceX also drew attention this week for reportedly forming special committees to discuss a possible merger. That remains an exploration, not a completed deal. Readers who want the corporate backdrop can follow our report on the SpaceX and Tesla merger committees.

What Muse Code and Muse Wonder Mean for Coding Tools

The part of Meta’s report that touches developers most directly is its model lineup. On October 12, Meta introduced Muse Wonder and Muse Wonder Full, which it calls its most capable models to date, and described the Muse Wonder models as offering frontier coding performance for building apps and games. Developers reach them through Muse Code and the Meta Model API. Muse Code is a coding agent that reportedly debuted in early August.

Pricing follows a ladder. Reported consumer plans for Muse Wonder access in the Muse app are $20 a month and $100 a month. Reported Muse Code plans for Muse Wonder Full access are $50 a month and $200 a month. Those Muse Code prices come from single-source coverage, so check Meta’s own pricing page before you budget around them. We broke down the top tier in our piece on Meta pricing Muse Code at $200 and mapped the field in how Muse Code takes on four AI coding rivals.

Mark Zuckerberg said Meta’s first enterprise monetization focus would be charging developers for access to proprietary foundation models and the Muse Code agent, according to earnings coverage. That is a notable shift. A company that built its reputation on open releases is now charging for its best coding model. It did release Muse Spark Open on October 15 as its most capable US open-weight model to date, so the open and paid tracks run side by side. Our write-up of day-0 SGLang support for Muse Spark Open shows how quickly the community picked up the open weights.

How Meta’s capex feeds the developer price list

The link between a $135 billion to $145 billion capex plan and a $200 monthly coding plan is direct. Inference on a frontier model consumes GPU time per token. Meta can lower its unit cost by building more capacity, or it can ration access with tiers and caps. It is doing both. The company reportedly reset usage for every Muse Code user after version 1.4.8 made the tool more token-efficient, which is a tidy example of efficiency work standing in for capacity.

Compute Rationing: Why Muse Is Limited to the US and Canada

Reports on October 21 said Muse availability was limited to the United States and Canada because of compute demands. That is the clearest example this quarter of capex directly constraining a product. A company with 3.64 billion daily users and an $82.45 billion cash pile still cannot serve a coding and assistant product everywhere at once. If you are an engineer outside North America, your access to Meta’s newest agent depends on how fast that $34.91 billion quarterly build converts into usable racks.

Tesla faces a different bottleneck. Its constraint is not serving tokens but producing hardware, which is why Optimus timing and Cybercab pre-production ramp costs feature in its Q3 discussion. Reports also named Megapack 3 and the Semi as products carrying pre-production ramp costs. The 69% jump in R&D reflects that the company is funding several programs at once.

What the Executives Said

The leaders of both companies framed the quarter in their own words, and the contrast is instructive. Mark Zuckerberg, Meta’s chief executive, said in the results statement, as Fortune reported: “Our investments are paying off, the core business is strong, and we’re seeing velocity across the board.”

Zuckerberg also pointed to the model launches. In a statement carried by CNBC’s live coverage, he said: “We hit major milestones this quarter toward delivering personal superintelligence for everyone – including launching Muse and Muse Wonder, our first frontier model.”

On the Tesla side, Elon Musk, chief executive of Tesla, discussed the capex revision on the earnings call. Per a call transcript on Webull, he said: “We’re revising our CapEx guidance to be approximately 20 billion for the current year as a result of managing our capital deployment efficiently and plan to fund it primarily from our cash flows from operations.”

On Optimus, Musk was candid about the gap between demos and utility. CNBC’s Tesla coverage quoted him: “There’s still quite a lot of work to do with regard to the AI and the robot for it to be truly useful for open-ended tasks.”

Read together, the four statements show two management styles. Zuckerberg sells spending as proof of progress. Musk sells restraint as discipline while admitting the product is unfinished.

Market Impact: Where the AI Trade Goes From Here

Investopedia flagged before the reports that Tesla and Meta results were due after the close, and the reaction suggests investors wanted evidence of returns, not only spending. A beat on revenue at Meta did not stop futures from sliding. The market is no longer paying for capex announcements alone. It wants to see monetization, such as the paid Muse tiers and developer access that Meta started to build in October.

Tesla’s lower capex can read two ways. Bulls see capital discipline and positive free cash flow of $1.531 billion. Bears see a company that planned to be an AI and robotics leader spending less on the infrastructure that goal requires. The after-hours dip of 2% to 3.3% shows that investors leaned toward the second reading, at least overnight.

For suppliers, the picture is cleaner. Meta’s raised floor on capex supports demand for chips, memory, networking and power equipment through the end of the year. Tesla’s cut removes one source of incremental demand, although its target of more training compute remains in place. The aggregate message to hardware vendors is mixed but net positive, as one hyperscale-sized spender committed more.

Historical Context: How Fast the Spending Curve Has Moved

A year ago, Tesla’s third-quarter capex was about $2.25 billion. This quarter it was about $4.73 billion, more than double. Its R&D expense rose from $1.63 billion to $2.76 billion over the same span. Meta’s trajectory is steeper. Its 2026 capex outlook of $135 billion to $145 billion exceeds what many companies spend on all operations in a year, and its Q3 capex alone ($34.91 billion) was 54% of the quarter’s revenue.

Meta’s Reality Labs shows the cost of earlier bets. The division’s 2026 losses are still projected at $18 billion to $19 billion, even as Meta says they should fall to around $15 billion next year. That history explains the market’s wariness. Investors have watched Meta pour money into one frontier before, and they now want proof that the AI shift will not repeat the pattern. The evidence so far is mixed but encouraging, since the Muse family has already produced paid plans, a coding agent and an open-weight model within about two months.

Competitive Comparison: Who Is Selling Coding Compute

Meta now competes for the same developer wallet as established coding assistants. All Things Muse reported that Meta’s newest model scored 1,726 points in its first Arena WebDev appearance, ahead of Claude Sonnet 5.5 and Gemini 4 Argon. That is a fan-site report of a leaderboard result, so treat it as indicative until Meta or the leaderboard confirms it. Our own comparison of Gemini 4 Argon and Claude Opus 5.5 on coding shows how tightly packed the top of the table already is.

Company	What developers can buy	Capex posture (Oct. 2026)	Status
Meta	Muse Code, Meta Model API, Muse Wonder Full	$135 billion to $145 billion for 2026	Live, US and Canada for Muse per reports
Meta (open track)	Muse Spark Open weights	Same budget	Released Oct. 15 per Meta Research
Tesla	No public coding or model product confirmed	About $20 billion for 2026	Compute for vehicles and Optimus
Anthropic	Claude Sonnet 5.5 (named in the Arena report)	Not part of this earnings cycle	Benchmark rival
Google	Gemini 4 Argon (named in the Arena report)	Not part of this earnings cycle	Benchmark rival

The comparison shows why this earnings cycle matters to the coding category. Meta is the only company in the two reports with a priced, shipping developer product. Tesla has the capital discipline but no developer offer. Anthropic and Google did not report this week, so their capex stories wait for later calendars.

Five Predictions for the Next Two Quarters

These are analysis, not reported facts, and each one hinges on figures already published.

Meta lands in the upper half of its capex range. With Q3 capex at $34.91 billion and a floor now at $135 billion, a fourth quarter near or above Q3 would put the full year toward the high end of $135 billion to $145 billion.
Muse access widens beyond the US and Canada. Compute is the stated reason for the limit, so new capacity should lift it. The timing depends on how fast racks come online.
Paid coding tiers multiply. Meta already runs a $50 and $200 ladder for Muse Code per reports. Expect rivals to answer with their own usage-based tiers instead of cutting list prices.
Tesla’s Q4 capex must jump to hit $20 billion. Nine-month spending of about $13.01 billion implies roughly $7 billion in the final quarter. A miss would show up as another guidance trim.
Investors will demand monetization data. Futures slid despite a Meta revenue beat, so the next report needs paid-subscriber or developer-revenue figures, not only larger budgets.

What Developers Should Do This Week

Engineers do not control Meta’s capex, but they do control how they spend their own tool budget. Start by pricing the options. Muse Code’s reported $50 and $200 plans give you a reference point against whatever coding assistant you pay for now. Check Meta’s official pricing page before committing, as the Muse Code numbers rest on a single source.

Next, plan for rationing. If a vendor limits a product by region or usage because of compute, your workflow should not depend on one provider. Keep a second assistant configured and keep prompts portable. Finally, track the cost side of your own AI use. Meta’s token-efficiency update and usage reset show that vendors tune consumption as hard as they grow capacity, and your bill can change without a visible price change. Our wider software coverage follows these shifts as they land.

Frequently Asked Questions

Why did US stock futures fall on October 22, 2026?

Futures fell after Meta and Tesla reported third-quarter results that raised fresh questions about AI spending. S&P 500 futures were down about 0.4% to 0.5%, Nasdaq-100 futures were down 0.7% and Dow futures were down 0.3%, according to CNBC and other market reports.

What did Meta report for the third quarter of 2026?

Meta reported revenue of $64.60 billion, up 26% year over year, against a consensus estimate of $63.37 billion. Earnings were $3.83 per share against an estimate of $3.57. Results include a $9.89 billion charge tied to an August 2026 settlement with US state attorneys general.

How much will Meta spend on capex in 2026?

Meta now expects capital expenditures, including principal payments on finance leases, of $135 billion to $145 billion. That narrows its earlier outlook of $130 billion to $145 billion. Third-quarter capex was $34.91 billion.

What did Tesla report and what is its 2026 capex plan?

Tesla reported revenue of $28.05 billion and adjusted earnings of $0.48 per share. It said capital expenditures would be approximately $20 billion for the year. Q3 capex was about $4.73 billion, below what analysts expected.

Did Tesla stock fall after the report?

Yes, in after-hours trading. CNBC reported a 2% decline, while Bloomberg-syndicated coverage reported 3.3%. The exact figure differs by timing, so no single number is confirmed.

What are Muse Wonder and Muse Code?

Muse Wonder and Muse Wonder Full are Meta’s most capable models, introduced on October 12, 2026. Muse Code is Meta’s coding agent, and the Meta Model API gives developers access to Muse Wonder. Reported consumer plans run $20 and $100 a month, and reported Muse Code plans run $50 and $200 a month.

Is Muse available outside the United States?

Reports on October 21 said Muse was limited to the United States and Canada because of compute demands. Meta has not given a date for a wider rollout in the coverage we reviewed.

Does Tesla sell any AI coding tool?

No coding model, API or developer product from Tesla has been confirmed. Its AI spending is aimed at vehicles, Optimus and training compute. Tesla said customer deliveries of Optimus could begin near the end of 2027, which is a target and not a firm date.

Meta Lifts Capex to $135B, Tesla Cuts to $20B [2026]