Rule of 70

Robert F. Smith's proposed successor to enterprise software's rule of 40, and his second category of investable software company: explicitly his own coinage, "what I call the rule of 70."¹

Explanation

The rule of 40 is the long-standing SaaS heuristic that revenue growth rate plus EBITDA margin should sum to at least 40: grow 40 percent at breakeven, or grow 10 percent at 30 percent margins, and either counts as a healthy business. It encodes a trade-off, buying growth with margin, where the sum is the efficient frontier.

Smith's claim is that generative AI moves the frontier itself, because it compresses every cost center simultaneously rather than trading one against another.

| Cost center | Mechanism | Stated effect | |---|---|---| | Product development | Agentic coding | Roughly 30 to 50 percent productivity gains writing new code; existing code somewhere between 2 and 12 percent | | Go-to-market | Agents as pre-sales, and in some cases sales | Customer acquisition cost comes down dramatically | | Services delivery | Agents replacing human professional-services effort | Services cost comes down | | Back office | General automation | Meaningfully more efficient |

The conclusion: "you'll be able to boost margins pretty dramatically, almost double them in many respects, for existing enterprise software companies that are serving large-scale industries."¹

The coding-productivity split

The most useful number on the page is a ratio rather than a single headline figure: 30 to 50 percent productivity gains on new code, 2 to 12 percent on existing code.¹ That gap is roughly an order of magnitude, and it is a far more honest picture than the greenfield-only figures usually quoted. The resolution offered is temporal rather than technical: "over time it's going to tip, more new code is going to replace the existing code," so the blended rate rises as a codebase turns over.¹ That is a real argument and an incomplete one, since it assumes rewriting beats maintaining, which is precisely the harder, contested claim in any debate about agent-maintained software; buyers who do not want to own the maintenance burden are a real brake on how fast that mix actually tips.

Why it matters

It is a valuation argument disguised as an operating metric. If the achievable frontier moves from 40 to 70, a company priced against the old frontier looks cheap, which turns a claim that enterprise software multiples have compressed into a coherent buy thesis rather than simple contrarianism; the rule of 70 is the numerator story that makes the multiple story work. It is also a margin thesis in a space where most AI-in-software commentary is a revenue thesis: capturing new revenue is one lever, but the same revenue at double the margin is a second, more controllable lever that a buyout owner can pursue directly. And it relocates the binding constraint on SaaS gross margin: classic SaaS margin was gated by human services delivery and human support, and if agents absorb those, gross margin approaches the software ideal and the residual cost of goods sold becomes inference, which is itself falling.

An existing high-efficiency benchmark helps calibrate the target: AppLovin had already reached a rule-of-40 score around 150, roughly 70 percent growth and 84 percent margins, before the agentic era. That a public company cleared more than twice Smith's stated target well before this shift suggests 70 is better read as a portfolio-wide aspiration than as a genuine frontier.

Tensions and open questions

It is a coinage, not yet a standard: no industry adoption, no cohort data, and no specific company named as having reached it. Simultaneity is asserted rather than demonstrated, since each of the four cost-center levers is individually plausible but all four landing at once, in the same year, inside a mature company, is a much stronger claim, and margin gains in software have historically been competed away rather than kept. Nothing addresses whether the gains are actually retained by the vendor rather than the buyer: if every enterprise software company gets the same cost compression from the same underlying infrastructure, the surplus tends to compete away to customers, and the rule of 40 simply becomes the new rule of 40 at a higher number for companies without a durable moat. Existing-code productivity sitting at 2 to 12 percent is a particular problem for a buyout portfolio specifically, since acquired enterprise software is mostly legacy code, and the "it will tip" answer is a bet on rewrite economics that remains genuinely disputed. No time horizon is attached to the promised margin doubling.