Sunil Venkataram
Back to garden
•
🌰 seedling
Grow 10 or Earn 40 - Two Paths for Mature Software Companies
type/noteinvestingsaasstrategy

Grow 10 or Earn 40 — Two Paths for Mature Software Companies

The thesis

The comfortable middle for software valuations is ending. Public markets have already delivered the first half of the message: growth rolled over, multiples compressed, and now the question is whether a company is going to earn a premium growth multiple or a fortress margin multiple. The companies that try to do neither will face persistent pressure from multiple compression, dilution, and stalled top-line.

David George (a16z) articulated this as the two-paths framework in March 2026. The framework is a forecast — a claim about what public markets will reward and punish over the next 12-24 months — not a guarantee. But the compression in software multiples through 2024-2025 is consistent with it, and the argument has internal coherence worth taking seriously.

Path one — Accelerate growth with AI-native products

The bar is higher than bolting chatbots or copilot UIs onto the existing SKU list. The bar is net-new products that can move the company’s total growth rate by 10 points within 12 months. Most incumbents don’t try this because it’s genuinely hard and requires rebuilding the company in parallel with building the new product. Specifically:

  • A vanguard team of ~5 people pulled from anywhere in the org. The bet is that in any large software company there are five people who can deliver 100x the value the current exec team thinks possible. The CEO’s first job is identifying them and handing them career-defining scope.
  • Process-capture sprints before product work. The vanguard’s first task is building a living context layer — not static PDFs — by harvesting SOPs, tickets, transcripts, CRM notes, support logs, and approval paths for every high-value workflow. This data is the raw material for the new AI products.
  • 50% of R&D redirected to net-new AI products within a month. Four-person pods with collapsed design/product/engineering, capped headcount (not compute), writing code day one.
  • Exec team refresh in month 1-2. The CEO watches which existing VPs can operate at the new tempo. The ones who can’t are replaced with vanguard members and AI-native up-and-comers. This is mandatory — the old exec team will be too slow.
  • Central engineering keeps the best engineers. Counterintuitive advice: don’t put your best engineers at the edges on net-new products. They stay in the core engineering org, reporting to the CTO, making sure the architecture can evolve. Put “people who ship and learn quickly” on the new products.
  • Business-model change is required. Seat-based pricing must give way to token/consumption pricing because that’s where net-new budget is flowing. If an agent can’t consume and pay for the product autonomously, the product isn’t ready for the new world.

Path one is a twelve-month death march. Most companies will fail at it. But the ones that succeed come out refounded: new leadership team, new product engine, growth reacceleration.

Path two — Rebuild to 40%+ true operating margins

The alternative for companies that cannot credibly accelerate growth: build the cash machine. The target is 40%+ true operating margins within 12-24 months, with stock-based compensation counted as a real expense (not carved out the way software has treated it for a decade).

Getting there demands deeper cuts than 10-20% headcount reductions:

  • Flattening management layers. Multi-level director/VP structures are incompatible with AI-era productivity gains.
  • Standardizing implementation and killing bespoke services. The professional services drag has to go.
  • Killing committees. Decision latency kills margin as much as headcount does.
  • Raising prices where pricing power exists. If the company owns a workflow or has switching cost, extract more from it. Move long-tail customers to higher floor pricing or let them churn.
  • Counting stock compensation as an owner-to-employee transfer. This is the single biggest conceptual shift, because it reframes what “profitable” means.
  • Increasing per-engineer token budget. Counterintuitively, the path to a smaller workforce starts with giving each remaining engineer much more compute budget. ~$1,000/month per engineer in token spend is table stakes; the 20-30x productivity stories from top operators are the ceiling, not the average.

The model is Hock Tan’s Broadcom — a public-market proof that radical cost discipline, product simplification, and price realization are possible at scale. It’s a harsh cultural template and not every founder can execute it, but the strong form exists.

Why the middle disappears

The forcing function is the treatment of stock-based compensation and the shift in customer budget allocation. Specifically:

  • SBC is now a real expense. Software’s free-cash-flow narrative of the last decade assumed dilution was free. Markets are repricing that assumption.
  • Customer AI savings hit the seat line first. When a customer deploys AI to make knowledge workers 30% more productive, the immediate budget effect is fewer seats. If a vendor’s entire model is seat-based, they’re standing in the shrinking side of the customer budget.
  • New customer budget is flowing to tokens, consumption, and outcomes. A vendor on the seat side of this line gets squeezed. A vendor on the consumption side gets the growth.

The middle — a seat-based vendor with moderate growth and moderate margins who hasn’t rebuilt either the product model or the cost structure — is the position markets will punish most.

The forcing question

From the article: “Founders should put one question on page one of every board deck: which path are we on? +10 points of revenue growth from net-new AI products? Or 40%+ true operating margins including SBC?”

The tell that a company hasn’t picked: evasion. “A little of both” or “we are evaluating options” is, per George, a signal that market pressure will continue. Picking badly is worse than picking well; picking nothing is the worst of the three outcomes.

How to use this framework

  • For public software equity analysis: put every holding on a forced two-path test. If a company has been “refactoring” for more than 12 months without visible progress on either axis, that’s a red flag.
  • For private company advising or board work: push the forcing question on page one of every board deck, as the article recommends. Make the evasive answer uncomfortable.
  • For founder/operator decisions: the vanguard-team move is transferable even to much smaller companies. The question “who are the five people who could 100x my output if I gave them the right scope” is worth asking at any size.

Honest caveats

  • This is a forecast, not a law. David George is talking his book as an a16z investor betting on new AI companies. The framework is sharpened for effect.
  • “12-18 months” is aspirational. Most real transformations take longer, and the companies that will succeed are those that start before the window closes, not those that start after it’s proven.
  • The middle will persist beyond 18 months. It will be compressed, not eliminated. The claim is about the direction of valuation pressure, not a binary outcome.
  • Path two survivors still need a thesis. Running for cash without a product thesis is a slow death, just at higher margins.
Backlinks (5)
Connected Notes