Meta Platforms trades at roughly 5.4 times projected 2027 sales, according to the CNBC Investing Club — about a third of the multiple the market pays for Anthropic, OpenAI or SpaceX.
Those three AI labs are unprofitable and not yet public. Meta already has a frontier-level model, more owned compute than either private lab, and a cash-generating advertising business underneath it, yet Wall Street is pricing almost none of that AI upside in.
"Surely, a model now proven to be in the running with those at the bleeding edge deserves some valuation greater than $0."
The call comes from the CNBC Investing Club team that manages Jim Cramer's Charitable Trust, which discloses long positions in both Meta and Alphabet.
I listened to the full segment so you can skip it.
Here are the 4 takeaways that matter.
👤 Author: Zev Fima, co-portfolio manager of the CNBC Investing Club, writing on behalf of Jim Cramer's Charitable Trust
📰 Published: 9 September 2026 on CNBC.com
🟣 Apple Podcasts | 📝 Full analysis | ⏱️ 3 min
Key Takeaways
Meta trades at about 5.4 times projected 2027 sales, roughly a third of what Anthropic, OpenAI or SpaceX command
Anthropic is valued at 15–30x sales, OpenAI at about 21x, SpaceX at about 20.6x
Muse Spark 1.3 already ranks among the top three frontier models, and the market is paying nothing for it
It scored 48 on the Artificial Analysis Intelligence Index, ahead of GPT-5.6 Sol, before OpenAI's Astra retook the lead
The new Muse agent app could out-distribute rivals through Meta's 3 billion-plus users across Facebook, Instagram and WhatsApp
Jim called it a "souped-up Alexa," and it launched a day before Apple's AI-powered Siri event
Meta could lease its idle Nvidia GPUs the way SpaceX does, but probably won't
Management likely sees more value using that compute for Meta's next model, Watermelon, than renting it to a rival like Anthropic
A cloud business alone could be worth about $745 a share, roughly $1.89 trillion in market cap
That's using Alphabet's roughly 22x forward earnings multiple, and doesn't count the AI lab's own value
Shares just reclaimed both key moving averages, and the MACD confirmed the move after the Muse Spark release
Four levels to watch: $650, $622, $599 and $550
The RSI is near the overbought threshold, but the team says that shouldn't scare anyone off the shares given the fundamentals
1. The AI Lab Worth $0
Shares jumped 5% Wednesday, though Meta is still down 2% year to date and needs to rise another 22% to reclaim last September's record high near $790. The move followed Meta's release of its Muse Spark 1.3 large language model on September 2, which the team says is reshaping how cheap the stock really is.
Even before any AI credit, Meta trades at about 19 times 2027 earnings estimates — roughly an "average" company multiple — while its core Family of Apps advertising business grows revenue 20% a year with strong operating cash flow
Muse Spark 1.3 scored 48 on the Artificial Analysis Intelligence Index, an independent industry benchmark, leapfrogging GPT-5.6 Sol, which had been OpenAI's leading model at the time
By Wednesday, OpenAI's new Astra model had retaken the lead over Muse Spark on that index, with Anthropic's Fable still on top — "Meta's AI lab is on the top three podium"
That Meta has built a frontier-level lab with more owned compute than either Anthropic or OpenAI "is simply not being appreciated by the market," the team wrote
The team sized the gap by comparing valuations: SpaceX, which owns xAI and its Grok chatbot, carries a market cap of about $2 trillion; Anthropic was valued at $965 billion in its latest private round and is reportedly seeking twice that in a planned IPO; OpenAI's latest round valued it at $852 billion. All three are expected to go public this year or in early 2027, and neither Anthropic nor OpenAI is believed to be profitable on a GAAP basis. SpaceX, also unprofitable today, is expected to turn profitable in the September and December quarters — largely on compute-lease deals it signed this year with Anthropic and Alphabet's Google
On a price-to-sales basis — used instead of price-to-earnings because none of the three AI labs is profitable — Anthropic trades at 15 to 30 times an estimated $65 billion annual revenue run rate, OpenAI at about 21 times a run rate topping $40 billion, and SpaceX at about 20.6 times projected fiscal 2027 sales of roughly $97 billion
Meta, with a market cap of about $1.65 trillion against 2027 sales estimates of about $305 billion, works out to a price-to-sales ratio of roughly 5.4 — "clearly, much lower than those other three" — while serving more than 3 billion users worldwide
"Surely, a model now proven to be in the running with those at the bleeding edge deserves some valuation greater than $0."
That's even before Watermelon, Meta's next model, which is "on the near-term horizon"
2. Muse vs Alexa and Siri
Meta released the Muse app on a limited basis Tuesday night — its own entry into AI personal agents, built on Muse Spark.
"It's getting rave reviews online, and having used it a bit ourselves, we concur that it's pretty incredible"
Jim described it as a "souped-up Alexa," referring to Amazon's voice assistant
xAI's GrokBot reached the market first, this summer, but Meta's more than 3 billion global users and Muse's integration with Facebook, Instagram and WhatsApp give it a much larger base to market to immediately
The release landed the day before Apple's iPhone launch event, where an AI-enhanced Siri was set to feature prominently — Muse, GrokBot and Siri are all now competing to be consumers' default personal assistant
3. The Cloud Card Unplayed
Meta has been viewed as the least attractive of the four big hyperscalers — Microsoft, Amazon and Alphabet are the other three — because it has no cloud-computing business to offset its aggressive data-center spending, which the team says has kept Meta's price-to-earnings ratio in the mid-to-low 20s, trailing its hyperscaler peers.
Jim has pushed Meta to build its own public cloud, and the team was disappointed the company said little about it on its late-July earnings call — but the option "isn't off the table"
Meta has stockpiled Nvidia GPUs and is developing its own custom AI silicon alongside Broadcom. The value of those chips "hasn't declined in the past month," and CoreWeave and others have shown the compute has a longer useful life than expected
If Meta wanted to lease spare GPUs to outside customers the way SpaceX does, the team believes it could "flip the switch and do just that" — though the revenue would matter far more to SpaceX's finances than to Meta's
With Watermelon on the way, management likely sees more long-term return in using that compute itself than renting it to a rival like Anthropic: "For that reason, we're honestly not sure whether Meta will ever go forward with a public cloud offering"
If Meta paired its ad business with a cloud arm the way Alphabet does — at Alphabet's roughly 22 times forward earnings — the stock would trade at about $745 a share, or roughly $1.89 trillion in market cap, which the team calls "arguably a very conservative price objective" since a cloud business would likely lift the earnings estimate behind that number too
That $1.89 trillion still wouldn't account for what could be a $1 trillion AI lab in the making, against Alphabet's current market cap of about $4.1 trillion
4. Four Levels, One Signal
"We consider ourselves to be fundamental analysts rather than technicians," the team wrote. But with Meta's chart and fundamentals now pointing the same way: "When the fundamentals and the technical setup are in agreement, then you may be looking at a very good opportunity to make some money."
Shares just reclaimed both the 50-day (about $599) and 200-day (about $623) moving averages, which now serve as support; a more cautious investor could wait for a pullback to the 200-day line to "retest" it
Overhead resistance, in a downtrend dating to August 2025, sits around $660
Below the 50-day average is "a bit of an air pocket," but a rising support level near $550 — in place since April, when AI-spending fears hit the stock — has held through the "Liberation Day" tariff selloff and the market decline at the start of the Iran war
The team's four levels to watch: $650 (Wednesday's intraday price), $622 (the 200-day moving average), $599 (the 50-day moving average) and $550 (the longer-term support)
The MACD confirmed the move: it crossed above its signal line below the zero line in late August, then crossed above the zero line itself in the days after the Muse Spark 1.3 release — both read as bullish
The Relative Strength Index sits around 70, right at the threshold where technicians call a stock overbought: "So, further strength from here in the near-term should be approached with caution, but given the fundamental updates, we don't think the current reading should scare anyone off the shares" — though the team cautions against being too aggressive here
Bonus Insights
Jim Cramer's Charitable Trust discloses long positions in both Meta and Alphabet, the two companies used in the sum-of-the-parts valuation comparison in Section 3
The team's bottom line is that Meta's advertising business alone already justified a market-multiple valuation, but a frontier AI lab and a hyperscaler-grade compute stockpile the market is pricing at close to zero is why they call this a stock to buy now, midterm volatility notwithstanding.
Products, Companies & Tools Mentioned
Meta Platforms (The subject of the analysis — its Muse Spark 1.3 model, Muse agent app and idle GPU stockpile are the three sources of value the team says the market isn't pricing)
Anthropic (Its Fable model leads the Artificial Analysis Intelligence Index; valued at $965 billion in its latest round and reportedly seeking double that in a planned IPO)
OpenAI (Its Astra model retook the top AI-benchmark spot from Meta this week; latest funding round valued it at $852 billion)
SpaceX and xAI's Grok (SpaceX's ~$2 trillion valuation and GPU-leasing business are the comparison the team uses to argue Meta could rent out its own idle compute)
Alphabet (Google's forward P/E of about 22 is the multiple the team applies to argue Meta could be worth $745 a share with a cloud business; also a SpaceX compute-lease customer)
Broadcom (Meta's partner in developing its own custom AI silicon, alongside its stockpile of Nvidia GPUs)
Nvidia (Maker of the GPUs Meta has stockpiled and could potentially lease out)
CoreWeave (Cited as evidence that GPU compute holds its value longer than previously assumed)
Artificial Analysis Intelligence Index (The independent benchmark on which Muse Spark 1.3 scored 48, briefly outranking OpenAI's prior flagship model)
Amazon's Alexa and Apple's Siri (Jim compared Muse to Alexa; Siri's AI overhaul was set to debut at Apple's iPhone event the day after Muse's launch)
If this was worth your time, send it to someone who follows the name.
Get the latest market chatter as it happens: