So Valve isn’t a B2B company per se, but it has evolved from a gaming platform to an epic marketplace (Steam) with unit economics that are almost unprecented. I still think we can learn fro it.
Valve will generate over $17 Billion in total gross revenue and $4 Billion in 90%+ gross margin Steam commission revenue in 2025 … with an estimated 330-360 employees. That’s $12+ million in revenue per employee. But here’s where it gets even more insane – if you look at just the ~79-person Steam team handling those commissions, that’s $50+ million per employee.
Forget about hot, lean AI startups. Even the most efficient hedge funds don’t touch these numbers.
I’ve spent 15+ years in B2B software, invested in 80+ companies, and I’ve never seen anything like this. So let’s break down what’s actually happening here and what it means for how we think about building efficient, capital-light businesses.
2025 Key Metrics:
- $16.2B+ in total Steam game sales (through mid-November, 44 days still remaining)
- $4B+ in Valve commission revenue
- 330-360 total employees (essentially unchanged since 2012)
- ~79 employees running the entire Steam platform
- $50M+ revenue per Steam employee
- 130M+ monthly active users
- 39.3M peak concurrent players (2024 record)
- 13,000+ games released annually
Valve Corporation is a privately-held gaming company founded in 1996 by former Microsoft employees Gabe Newell and Mike Harrington. While most people know Valve for iconic games like Half-Life, Counter-Strike, Portal, and Dota 2, the company’s real money-maker is Steam – the dominant PC gaming platform that controls approximately 70% of digital PC game distribution worldwide.
Think of Steam as the “AWS of gaming” – it’s a digital storefront, distribution platform, community hub, and payment processor all rolled into one. Developers upload their games, Steam handles everything else: downloads, updates, multiplayer infrastructure, anti-cheat, payment processing, and discovery for 130M+ monthly active users.
Here’s what makes Valve remarkable: They’ve built one of the most profitable businesses in tech with an almost comically small team. No VC funding. No board. No managers. Just ~360 people generating $6.5B+ in annual revenue. It’s the closest thing to a “perfect” platform business model I’ve ever studied.
Learning #1: The 30% Rule – Platform Tax vs. Platform Value
The Metric That Matters: Valve takes a 30% commission on game sales up to $10M, 25% from $10-50M, and 20% above $50M. Steam generated over $4B in commissions through November 2025, up from $3.2B in 2024 and ~$100M in 2009 – a 40x increase in 16 years. Total game sales on Steam exceeded $16.2 billion in 2025 (with 44 days still remaining in the year).
Here’s what’s fascinating: Everyone argues about whether 30% is “fair.” Epic launched with 12% to compete. Developers complain. Regulators investigate. But the number that matters isn’t the percentage – it’s the value created per dollar of commission.
Steam isn’t just distribution. It’s:
- Discovery engine (13,000+ games released annually)
- Automatic updates and cloud saves
- Community features and workshop
- Anti-cheat and matchmaking infrastructure
- Built-in payment processing for 130M+ MAU (2025)
The brutal truth? Most platforms that charge 30% don’t deliver 30% of value. Valve does. That’s why despite Epic’s 12% offer and years of competition, Steam still controls ~70% of PC game downloads.
The SaaStr Parallel: In B2B SaaS, we obsess over take rates in marketplaces. The magic number is usually 10-25%. But Valve proves that if you’re genuinely creating massive value – real distribution, real discovery, real infrastructure – customers will pay premium rates. The key is that the platform has to be so valuable that opting out costs more than opting in.
Tactical Takeaway: Don’t compete on price alone. Compete on being so embedded in your customers’ workflow that switching costs exceed your commission. Valve built tools, community features, and player accounts so deep that games essentially require Steam to succeed on PC.
Learning #2: The 79-Person Platform – Team Size as Strategy
The Brutal Metric: Approximately 79 employees still run Steam in 2025 while processing 13,000+ game releases annually and serving 130M+ monthly active users. That’s 1.6M+ users per employee. Meanwhile, Valve’s total headcount has remained remarkably stable at 330-360 employees (virtually unchanged since 2021), including ~180 in game development and ~40 in hardware.
This isn’t just lean. This is a fundamentally different operating model.
Compare to similar-scale platforms:
- PlayStation: ~$2M revenue per employee
- Nintendo: ~$1.5M revenue per employee
- Xbox: ~$1.1M revenue per employee
Valve’s 45-50x advantage (at $50M+ per Steam employee in 2025) isn’t just efficiency. It’s architectural.
Here’s what I’ve learned from this: Valve designed Steam from day one to be maximally automated and self-service. Publishers upload games. The algorithm handles discovery. The community moderates itself. Updates push automatically. The payment infrastructure scales infinitely.
This is the opposite of how most B2B companies scale. We add CSMs. We build implementation teams. We create support tiers. Every $1M in new ARR = 3-5 more people.
The SaaStr Parallel: The best B2B companies I’ve invested in have NDR of 120-130% with almost no humans involved. The mediocre ones require armies of CSMs to achieve 110%. The difference? Product-led growth engines that handle onboarding, expansion, and retention automatically.
Owner.com (a portfolio company showing 10.63x MOIC) gets this. They built AI-powered tools that handle 80% of what used to require sales calls. RevenueCat (6.44x MOIC) built such a good self-service product that developers integrate without talking to anyone.
Tactical Takeaway: Every time you’re about to hire someone, ask: “Could we build a feature instead?” Valve’s entire competitive moat is that they built systems where competitors hired people. Systems scale. People don’t.
Learning #3: Gross Margins Above 95% – The Ultimate Unit Economics
The Numbers That Matter: Steam’s operating margins remain near 60% by 2025, with estimated operating profit approaching $2.4B on over $4B in commissions. But here’s the kicker – the gross margins on digital distribution are effectively 95%+.
Why? Zero cost of goods sold. No inventory. No manufacturing. No fulfillment centers. No returns. A game selling 1 copy costs Valve the same as a game selling 10 million copies. The incremental cost approaches zero.
Let me spell out what this means for capital efficiency:
- Valve is bootstrapped (never raised VC)
- $6.5B+ estimated annual revenue by 2025
- ~40-50% net profit margin ($2.5B+ in profit)
- 330-360 employees
- Average salary of $1.3M+ per employee (2021 data, likely higher now)
This is the dream, right? But here’s what everyone misses – you can’t retrofit this level of efficiency. Valve built for it from the beginning.
They made critical early decisions:
- Digital-only when physical retail still dominated (2003)
- Platform-first instead of just game sales
- Developer tools that made publishers self-sufficient
- Community-driven discovery instead of editorial curation
Each decision optimized for margin over market share initially. But over 20+ years, those margins compounded into the most profitable business model in gaming.
The SaaStr Parallel: The difference between 70% gross margins and 85% gross margins is the difference between a decent SaaS business and a generational one. That 15 points determines:
- How much you can spend on CAC
- How fast you can grow without dilution
- Whether you need a Series C, D, E to scale
- Exit multiples (10x vs. 20x revenue)
I see too many founders accepting 60-70% gross margins as “good enough.” It’s not. Gorgias (8.35x MOIC in my portfolio) obsessed over getting to 85%+ margins before aggressive scaling. That discipline creates options.
Tactical Takeaway: Every 5 points of gross margin matters exponentially at scale. Fight for them early when it doesn’t feel like it matters. At $100M revenue, 5 points = $5M more profit = 10-15 more people or 30-40% more go-to-market spend without raising dilutive capital.
Learning #4: Flat Structure Enables Speed at Scale – Organization as Product
The Organizational Metric: Valve operates with zero managers, no hierarchy, and employees literally move their desks to work on projects they choose. With 330-360 people generating $6.5B+ in revenue, that’s a 0% management overhead model. Remarkably, this headcount has been virtually unchanged since 2012 (when they had ~250 people), even as revenue has grown 3-4x.
Most people hear this and think “that would never work for us.” Here’s what they’re missing: Valve’s structure isn’t about being nice or employee-friendly. It’s about maximizing decision velocity.
Think about typical B2B software companies at $50-500M revenue:
- 5-8 layers of management
- Decisions require 3-4 meetings minimum
- Feature prioritization takes weeks of debate
- Resource allocation is political, not optimal
Now think about Valve’s model:
- Engineers see opportunity in Steam economy
- Team self-assembles around it
- Ships to 40M+ daily users (2024: 39.3M peak concurrent players)
- Iterates based on direct feedback
The speed difference is 10x minimum. But here’s the catch – this only works if:
- You hire perfectly (Valve’s bar is famously absurd)
- Compensation is transparent (everyone knows the stack ranking)
- Product is the organizing principle (not titles or territory)
- You’re willing to fire non-performers instantly (no place to hide)
Yanis Varoufakis (yes, the economist) joined Valve in 2012 and wrote about their structure. His key insight: “Traditional firms exist to solve coordination problems. Valve eliminated those problems through product design.”
The SaaStr Parallel: The fastest-moving companies I know don’t have fewer meetings because of discipline – they have fewer meetings because they built systems that eliminate the need for meetings.
At SaaStr Fund I, my best deals came from founders who moved faster than the market. Pipedrive, Algolia, Talkdesk – all perfect records. Speed was the common thread. Not recklessness – systematic velocity.
When I look at my portfolio companies with 5-10x MOICs, they all have “small company speed” despite being at scale. The ones grinding along at 2-3x? Too many managers. Too many committees.
Tactical Takeaway: Count your management layers. If you’re at $10M ARR with more than 3 layers, you’re building bureaucracy, not a business. At $50M ARR with more than 4 layers, you’re already too slow. Every layer halves your decision speed.
Learning #5: Network Effects at the Asset Layer – Building Moats Where Others See Features
The Data Point: Steam Community Market processes millions of transactions monthly for in-game items (skins, weapons, etc.) in games like CS2 and Dota 2. Valve takes a small cut of each transaction. This isn’t revenue – it’s a moat.
Here’s what genius looks like: Most platforms think of network effects at the user layer. Valve built them at the asset layer.
Every CS2 skin, every Dota 2 item, every trading card exists only within Steam. Players spend thousands of hours and real money accumulating these digital assets. Try to get them to leave Steam? They’d be abandoning their inventory.
The numbers:
- CS2 skin market exceeded $1B in transactions annually
- Some rare items sell for $100,000+
- Players have “Steam accounts” worth $5,000-50,000 in assets
This creates what I call “economic gravity” – the more you use the platform, the more expensive it becomes to leave. Not because of contracts or lock-in, but because of value accumulated within the ecosystem.
The Numbers Behind the Moat:
- 1 billion+ registered Steam accounts (lifetime)
- 130M+ monthly active users (2025)
- 39.3M peak concurrent users (2024 record)
- Average user has 100+ hours invested in library
- Plus digital assets worth real money
Compare to Epic Games Store launching with 12% commission (vs. Steam’s 30%): Despite 6+ years of aggressive subsidies and exclusives, Epic captured maybe 10-15% market share. Why? Because Epic is just distribution. Steam is economic infrastructure.
The SaaStr Parallel: The best B2B companies don’t just have high switching costs – they have accumulating value that makes switching economically irrational.
Salesforce doesn’t just have your CRM data. They have:
- 10 years of historical records
- Custom workflows that run your business
- Integrations with 100+ other tools
- Reports executives view daily
- User training and muscle memory
Ripping that out doesn’t just cost money – it risks the entire go-to-market engine.
In my portfolio, RevenueCat gets this perfectly. They don’t just process app subscriptions – they become the source of truth for subscription data, analytics, and experimentation. After 6-12 months, you’ve built dashboards, attribution models, and pricing experiments all dependent on RevenueCat’s data. Moving to a competitor means losing historical continuity.
Tactical Takeaway: Ask yourself: “If our customer used our product for 2 years, what have they accumulated that makes leaving painful?” If the answer is “our contract,” you don’t have a moat. Build features that create accumulating value: analytics history, workflow automation, team collaboration artifacts, or literal economic assets.
The Meta-Lesson: Efficiency at Scale Requires Different DNA
Here’s what I’ve learned studying Valve against 15+ years in B2B SaaS:
You can’t hack your way to Valve-level efficiency. This isn’t about fractional CTOs or lean operations. This is about fundamentally different architectural decisions made when efficiency seemed like a disadvantage.
In 2003, Steam was worse than retail distribution on almost every metric:
- Slower downloads (vs. buying a CD)
- Required internet connection (not universal)
- No used game market
- No lending to friends
- Required running background software
Valve chose digital distribution not because it was better for users, but because it would eventually create dramatically better unit economics. They were willing to be worse initially to be orders of magnitude better eventually.
That’s the pattern I see in the best B2B companies too. They make architectural choices that:
- Enable 85-90% gross margins (not 70%)
- Design for self-service (not high-touch)
- Build systems (not hire people)
- Create network effects (not just stickiness)
- Optimize for terminal value (not current convenience)
And the results speak for themselves: Valve’s employee count has been essentially flat at 330-360 people for over a decade, while revenue has grown from ~$2B to $6.5B+. That’s the compounding power of the right architectural decisions.
What This Means for Your B2B Company
If you’re building B2B software today, here’s what Valve’s model teaches:
At $1-10M ARR: Make the hard architectural decisions NOW. Choose self-service over high-touch, even if it slows initial growth. Build for 85%+ gross margins, even if services revenue is tempting. Design systems that eliminate the need for future headcount.
At $10-50M ARR: Measure efficiency relentlessly. What’s your revenue per employee? It should be $300K-500K minimum, ideally $500K-750K. Count management layers – you shouldn’t have more than 3. Any decision taking >1 week should be escalated as a process problem.
At $50M+ ARR: You either built the foundation for Valve-style efficiency, or you didn’t. If you didn’t, you’re now in a different category – you can still build a great business, but the multiple at exit will reflect your need for capital and headcount to scale. If you did, you’re building something potentially generational.
Short-Term Sacrifice for Long Term Dominance
The real lesson from Valve isn’t that we should all operate with flat hierarchies or generate $50M per employee. It’s that the decisions that create extraordinary efficiency are usually invisible and uncomfortable when you’re small.
Valve spent 2 years building Steam while competitors shipped games faster. They built community features while competitors focused on transactions. They hired 79 exceptional people for Steam while competitors hired hundreds. And critically, they’ve maintained that discipline – keeping headcount essentially flat for over a decade while revenue tripled.
In B2B SaaS, I see the same pattern. The companies generating 5-10x returns made hard, expensive, non-obvious choices early:
- 6 months building self-service onboarding (vs. doing it manually)
- Turning down enterprise customers to protect product simplicity
- Investing in APIs when no one was asking for them
- Building analytics infrastructure before $10M ARR
These decisions feel wrong when you’re optimizing for next quarter’s ARR. They’re exactly right when you’re building for $100M+ revenue and a generational outcome.
Valve proves the ultimate arbitrage in tech: short-term sacrifice for long-term structure. Build the foundation for efficiency when it seems expensive and unnecessary. Sixteen years later, you’re printing $4B+ in commission revenue with ~80 people running the core platform and total headcount still under 400.
That’s the dream.
Jason Lemkin is the founder of SaaStr and managing partner of SaaStr Fund ($200M), with investments in companies like Owner, Assistant UI, Alloy. Algolia, Pipedrive, Talkdesk, Salesloft, Gorgias, and RevenueCat.
