M&A pricing in tech can be a profoundly weird thing. And sometimes, understanding the weirdness is the key to understanding how exits actually work.
Meta Buys Manus for $2B+ at $125M ARR
Meta just acquired Manus, the AI agent startup, for reportedly over $2 billion. At acquisition, Manus was running at ~$125M ARR. That’s roughly a 16x multiple.
But here’s the thing: that $125M ARR is a rounding error on Meta’s $150B+ in annual revenue. It’s immaterial. It’s a distraction, operationally.
What will likely happen: the team gets absorbed into Meta AI, the product gets integrated across Facebook, Instagram, and WhatsApp, and the standalone subscription slowly evaporates. Whatever Meta does with Manus, they likely won’t be investing more in it as a standalone product.
You sell your startup — in part, sometimes at least — on ARR multiples that in the end don’t really matter, because the ARR is immaterial or abandoned by the acquirer.
It’s just how it works.
But It Gets Weirder: The Deals Where Revenue Actively Walks Away
Manus is almost a clean case. The revenue is small, it gets absorbed, life goes on.
But look at Scale AI. Meta paid $14.3 billion for a 49% stake in a company running at ~$1.5B ARR. That’s real revenue. That matters.
Except… Meta’s investment immediately triggered customer flight. OpenAI cut ties. Google — which was planning to spend $200M with Scale in 2025 — pulled back. Microsoft and xAI reportedly started exploring alternatives.
Meta paid $14.3 billion for a chunk of a company, and a massive portion of its revenue started walking out the door the moment the deal closed.
The revenue wasn’t acquired. It was actively damaged.
And yet — the deal still made sense to Meta. Because the revenue was never the point.
Nvidia “Licenses” Groq for $20B
Same pattern, different structure.
Nvidia paid $20 billion for Groq’s IP and key talent — including founder Jonathan Ross, who created Google’s TPUs — in what they called a “non-exclusive licensing agreement.”
Groq was valued at $6.9 billion just three months earlier. They were targeting $500M in revenue. They had a differentiated LPU architecture that was genuinely threatening Nvidia in the inference market.
Now? Groq “continues to operate” with its CFO as CEO. The founder, president, and core engineering team all joined Nvidia. The technology gets folded into Nvidia’s architecture.
Translation: the company technically exists. But everything that made Groq Groq now belongs to Nvidia. The revenue stream is orphaned.
This Is the New Playbook — For Some Deals Only
These aren’t anomalies. This is how AI M&A works now. For certain IP and talent driven deals, at least.
Microsoft paid $650M in “licensing fees” that included hiring Inflection’s CEO as their top AI executive. Amazon hired away Adept AI’s founders. Nvidia did the same thing with Enfabrica earlier this year for $900M.
The deals are structured as “licensing agreements” or “acqui-hires” to avoid regulatory scrutiny. One analyst called the Nvidia-Groq structure a way to keep “the fiction of competition alive.”
But the economics are the same as traditional M&A. Massive valuations. Justified by revenue. With that revenue largely orphaned or destroyed post-close.
Why This Happens
In strategic M&A, the revenue is proof — not product.
- The ARR demonstrates execution. Manus’s $125M proved product-market fit. Scale’s $1.5B proved enterprise traction. Groq’s trajectory proved the technology worked. But the acquirer isn’t buying the revenue stream.
- They’re buying optionality and positioning. What can Meta now build with Manus’s agent technology? What can Nvidia now do with Groq’s inference IP that competitors can’t? That’s worth far more than the present value of any revenue stream.
- Integration means transformation. When you get absorbed into a $150B company, your product doesn’t stay your product. It becomes a feature, a capability, IP to be integrated. The standalone business is almost always deprioritized.
- Sometimes they’re buying you so no one else can. Nvidia buying Groq isn’t just about gaining inference capabilities — it’s about ensuring AMD, Google, or Amazon don’t get them.
What This Means for Founders
If you’re building something strategic acquirers want:
- Your ARR is your proof, not your product. Traction demonstrates you’ve built something real. But the acquirer is buying what you represent to their strategy, not what you are today.
- The “licensing deal” is the new acquisition. Especially in AI, expect deals structured to avoid regulatory scrutiny while achieving the same outcome. Your exit may not look like a traditional acquisition — but the check clears the same.
- Don’t over-optimize for ARR quality in late-stage negotiations. Yes, you need real customers and real revenue. But the strategic premium often has little to do with NRR or gross margins. It has to do with how badly the acquirer needs what you’ve built.
- The revenue may not survive. And that’s okay. Your equity outcome is locked at close. What happens to your product after is the acquirer’s problem.
In Contrast: PE Deals Are the Exact Opposite
Everything I just described? Private equity deals work the opposite way.
In a PE acquisition, the revenue is the product. It’s the entire point.
Look at the recent Thoma Bravo deals:
- Dayforce: $12.3 billion take-private of the HCM platform — their largest deal ever. They’re buying ~$1.5B in recurring revenue that they plan to grow, optimize, and eventually exit. The product stays. The customers stay. The revenue is the asset.
- SolarWinds: Taken private (again) by Turn/River for $4.4 billion, after Thoma Bravo and Silver Lake owned it for years. Same playbook: operational improvements, margin expansion, hold, exit.
- PROS Holdings: $1.4 billion take-private, with Thoma Bravo planning to combine the B2B business with their existing portfolio company Conga. Classic roll-up — consolidate revenue streams, improve margins, create a larger asset.
- SailPoint: Bought for $6.9 billion, optimized as a private company, then taken public again in early 2025 at a $10B+ valuation. The PE playbook executed perfectly.
In every case: buy the ARR, improve the operations, expand the margins, sell the company. The revenue doesn’t get orphaned — it gets optimized.
That’s why PE deals trade at lower multiples than strategic deals. A PE firm paying 8-12x ARR is pricing in what they can actually do with that revenue — grow it 15-20% annually, expand margins, hold for 5-7 years, and exit at a similar or slightly higher multiple.
A strategic acquirer paying 16x for Manus or 3x a recent valuation for Groq? They’re pricing in something else entirely. Something that has almost nothing to do with the revenue itself.
Two Completely Different Games
So when you’re thinking about your exit, understand which game you’re playing:
- PE exit: Your ARR quality is everything. NRR, gross margins, growth efficiency, customer concentration — it all gets scrutinized because it all gets kept. The revenue is the asset.
- Strategic exit: Your ARR is proof of concept. What matters is how badly the acquirer needs your technology, your team, or your competitive position. The revenue might not survive the acquisition — and that’s fine.
Both can be great outcomes. But they’re completely different math.
M&A Math in Tech Can Be Very Strange. And Very Context Sensitive.
The M&A math in tech is strange. Revenue multiples in strategic deals are really just a convenient fiction — a way to justify a number that reflects something far messier.
Manus’s $125M ARR got them to $2B+. Scale’s $1.5B ARR got them a $29B valuation — and then started evaporating. Groq’s revenue trajectory got them to $20B in a deal that orphaned the product entirely.
Meanwhile, a PE firm would have looked at those same companies and asked very different questions. Questions about the revenue itself — not what it represents.
That’s not a bug. That’s two completely different ways to value a company. And in 2025, understanding which game you’re playing might be the most important strategic decision you make.
