Only Two Paths Left in Software—No Middle Ground?
Deep thoughts on AI and aspirations —— ByteDance Deep Thinking Circle
A harsh verdict has been circulating in the software industry lately: the middle ground has disappeared, leaving only two paths. Either build genuinely novel AI-native products and boost your company’s overall revenue growth rate by 10 percentage points or more within 12 to 18 months; or restructure your organization to achieve a true operating margin of 40%—ideally over 50%—after deducting stock-based compensation. Companies stuck between high growth and high profitability have no path forward.
This judgment deserves serious attention. It essentially declares a death sentence for a certain way of operating: software companies with mediocre growth, mediocre profits, maintaining valuations through narrative alone.
How the Middle Ground Collapsed
Over the past decade, software companies grew accustomed to a playbook: when growth slowed, talk profits; when profits looked bad, talk growth; when neither stood out, talk ecosystem and stickiness. Capital markets once bought in because software stocks offered a certainty premium. Now this logic has been dismantled by two developments.
One is that growth is no longer scarce. After AI-native products drove down delivery costs for certain workflows, customers began reassessing software budgets. Companies unable to sustain high growth lose the growth stock premium. The other development is more subtle: stock-based compensation is finally being counted as real money. Software companies used to tout free cash flow margins, excluding stock compensation as if dilution weren’t a cost. Now the market treats every share issuance as wealth transfer from shareholders to employees, and true margins after deduction look far worse than reported.
These two factors combined caused the middle ground to collapse. Companies claiming neither pole face only valuation compression and continued dilution.
Why You Can Only Choose One Path
Twelve to 18 months is only enough to walk one path—superficially a resource problem, fundamentally an organizational one.
Taking the growth path requires entering wartime mode. A repeatedly validated approach: first identify the undervalued core talent in your organization, disregard their titles, and give them frontline assignments. Conduct an information inventory around high-value workflows, gathering materials like SOPs, tickets, meeting minutes, requirement docs, and support logs to build a dynamically updated context layer rather than piling up static documents. Then use a one-month observation period to determine executive retention: who’s supporting this elite squad and who’s watching from the sidelines becomes crystal clear. Have exit conversations quickly with executives falling behind; give their positions to those who won the information war and proven AI backbone members.
Next, redirect half your R&D resources toward pure AI new business. The organizational approach matters: four-person strike teams, design-product-engineering fused into a single combat unit writing code from day one; cap team headcount strictly, set no limits on compute budgets. Reserve the best time for product managers closest to customers; keep top engineers in the central engineering department guarding the technical foundation rather than scattering them across peripheral squads—otherwise fragmented tech stacks become debt that takes years to repay. This approach’s underlying assumption: AI has raised individual engineer output ceilings dramatically, while ten-person decision layer communication costs consume most gains. So cap headcount, fund budgets, let decisions concentrate upward rapidly, with executives clearing obstacles for frontline teams with full days weekly.
The other path transforms yourself into a profit margin machine: within 12 to 24 months, achieve 40% operating margin after deducting stock-based compensation, ideally 50%. Cutting 10% of staff won’t get you there. It requires reducing management layers, standardizing delivery, compressing custom services, raising prices where you have pricing power, moving long-tail customers to more expensive tiers or simply abandoning them. Broadcom provides a ready reference: when Hock Tan took over, the company was losing over $200 million annually; within three years, headcount dropped from 6,500 to 3,600, unprofitable business lines were cut, gross margin floors enforced, ultimately creating a cash flow superpower. This path works, but the cost is brutal—most founders can’t replicate it.
You must also honestly face a prerequisite: old moats are losing effectiveness. Data alone isn’t enough; system integration’s replication barriers are falling; when agents can operate across systems, workflow and interface advantages depreciate, data migration becomes easier. Companies taking the profit margin path must first assess which moats remain and concentrate resources on core advantages that still command pricing power and retention.
How to Choose—Look at Your Chips, Not Your Posture
Neither path is superior—only appropriate or not. Taking growth requires high-value workflows, mineable internal talent, channels close to customers, and cash to survive a 12-month sprint. Taking profit requires core modules with pricing power and the resolve to restructure the organization. The most dangerous answer is “we’re pursuing both”—strategic ambiguity’s cost is continued valuation compression. This judgment deserves the first page of every board deck: which path are we actually taking? The answer needn’t be perfect, but it must exist.
Don’t Mistake “Wait and See” for a Third Path
The most common self-rescue for companies stuck in the middle is announcing “stabilize the base first, then charge when winds favor us.” This posture sounds prudent but often just surrenders judgment to time, which only continues compressing valuations and accelerating dilution. The truly honest question is: does our cash support reaching a validation point on either path?
If not, only two options remain. Actively pursue M&A channels, packaging team and product for companies willing to pursue growth; or immediately contract to your smallest core, pressing all resources into the business segment that still has pricing power, making it cash-generative. M&A in software has never been a shameful outcome—it’s simply the normal exit for middle-ground companies after valuation logic shifts. The fear isn’t choosing wrong, but refusing to choose while unable to stop, letting the company become a frog in warming water, burning through cash until forced to decide at zero.
One final boundary. Accelerating growth by 10 percentage points within 12 to 18 months is a high bar for most companies; failing to meet it doesn’t mean the company lacks value—it only means old valuation methods no longer apply. The true decision starting point is assessing your workflows, customers, people, and cash in hand, not rushing to announce which path you’re taking.