Sam Altman has observed that both startups and large companies face unique struggles during the current wave of AI disruption, while firms that already have product-market fit often adapt more effectively (OfficeChai). Startups, despite their speed, often lack the foundation to scale. Giants, despite their resources, get trapped in bureaucracy. Companies with strong user adoption and proven fit, on the other hand, can use AI to deepen their advantage.

This raises an important question for product leaders: why does AI disruption play out so differently depending on the size and maturity of a company?

To answer it, we can turn to three proven frameworks. Hamilton Helmer’s 7 Powers shows how competitive advantage evolves. Porter’s Five Forces explains how market structures shift. And the Capability Stack highlights where AI creates or erodes leverage. Together, they reveal the hidden forces shaping winners and losers in the AI era.

7 Powers in the Age of AI

Hamilton Helmer’s book 7 Powers describes durable sources of strategic advantage. There are seven in total: scale economies, network effects, counter-positioning, switching costs, branding, cornered resource, and process power.

Here we’ll focus on the four most relevant in the AI context: scale economies, network effects, process power, and counter-positioning. The others—switching costs, branding, and cornered resource—still play a role but are less central to explaining the disruption dynamics we’re seeing today. For example, switching costs remain low in most AI tools, branding often shows up as part of trust, and cornered resources like data or GPUs reinforce dynamics we’ll cover elsewhere.

Scale Economies
Cloud hyperscalers such as Microsoft and AWS benefit disproportionately from AI. Their capital investments in data centers and GPUs create cost curves that startups cannot match. For example, Microsoft’s Azure revenue continues to accelerate as enterprises adopt AI services layered on top of its infrastructure (The Outpost). Startups, by contrast, face rising costs to train or fine-tune models when they must rent scarce GPUs.

Network Effects
AI tools with broad adoption compound their lead. GitHub Copilot is a clear example. More than 77,000 organizations adopted Copilot in 2024, up 180 percent year over year. Microsoft said Copilot now accounts for over 40 percent of GitHub’s revenue growth. This kind of usage feeds into better training signals, more integrations, and deeper stickiness—a classic flywheel. Competing startups in code generation face an uphill battle because they cannot replicate that network effect.

Process Power
Startups still hold an advantage in how quickly they can experiment and adapt. The pace matters because AI technology itself is advancing every few months. A controlled study found that developers using Copilot completed a programming task 55.8% faster than those without. The organizations that learn fastest, not just the ones with the largest datasets, will enjoy process-based advantages. Yet, for large firms, governance often slows decision-making to an annual cadence—mismatched to AI’s quarterly leaps.

Counter-Positioning
AI-native startups also use counter-positioning to attack incumbents. When OpenAI launched ChatGPT, it directly challenged Google’s search-based advertising model. Google hesitated to release similar functionality quickly, knowing it could cannibalize ad revenue. That hesitation gave OpenAI space to capture mindshare.

Taken together, the 7 Powers show why startups face steep cost disadvantages but may still find opportunities through speed and counter-positioning. Giants can rely on scale and distribution, but their process power is often weak. Firms with existing product market fit, especially those already embedded in workflows, can stack multiple powers to defend and grow.

Porter’s Five Forces Rewired by AI

Michael Porter’s classic framework helps explain how AI reshapes the competitive landscape. Each force is in flux.

Threat of New Entrants
AI has lowered barriers to entry. With open-source models and cloud APIs, small teams can build products that look sophisticated on the surface. As of 2024, over 14,000 AI startups were active globally, fueled by venture funding that reached $100 billion. The flood of entrants makes differentiation harder and increases noise in the market.

Supplier Power
Suppliers now hold extraordinary leverage. NVIDIA’s near monopoly on advanced GPUs has created bottlenecks and pricing power (NYT). Similarly, model providers like OpenAI, Anthropic, and Google set terms for access to frontier models. Startups dependent on them may find margins squeezed.

Buyer Power
Customers are more demanding. AI is no longer a novelty—it is expected (or soon will be). Every SaaS vendor feels pressure to add AI assistants, copilots, or personalization features. Buyers can easily switch if one vendor’s AI offering feels weaker. This dynamic increases buyer power and compresses the differentiation window.

Threat of Substitutes
Substitutes appear overnight. Jasper AI was an early leader in AI copywriting. But as soon as Canva, HubSpot, and Notion embedded generative writing features into existing workflows, Jasper’s growth slowed. For startups, the risk is being substituted by incumbents who integrate AI faster into products that customers already use.

Industry Rivalry
Rivalry is intense and accelerating. McKinsey reported that enterprise AI adoption jumped from 20% in 2017 to 55% in 2023. (Although I am skeptical of that higher number in ‘23, the adoption rates are rising from our own experiences.) With so many entrants and rising buyer expectations, the fight for market share is fierce.

Porter’s model highlights why both startups and giants feel squeezed. Startups face supplier power and substitution risk. Giants face buyer pressure and rivalry from new entrants. The companies with proven adoption and distribution are better equipped to navigate this turbulence.

The Capability Stack

A third way to view advantage in AI is as a capability stack. Competitive edge comes from controlling more layers of the stack: data, distribution, workflow integration, and trust.

Data
Proprietary datasets give companies leverage. BloombergGPT, trained on decades of financial data, can answer questions that generic LLMs cannot. Similarly, medical AI startups with access to labeled clinical datasets create moats that general-purpose models lack.

Distribution
Companies with wide distribution can roll out AI features at scale instantly. Microsoft introduced Copilot across Microsoft 365, giving hundreds of millions of workers access. Adoption is not only rapid but also baked into existing enterprise budgets.

Workflow Integration
Being embedded in workflows is a defensible position. Figma’s AI design tools are not standalone—they sit within a design ecosystem already used by teams daily. Startups trying to sell point AI design apps face high switching costs.

Trust and Safety
Trust is increasingly a differentiator. Enterprises care about compliance, copyright, and data security. Adobe positioned Firefly as “safe for commercial use” because it trained only on licensed data. In contrast, tools like MidJourney face uncertainty around intellectual property, which makes enterprises cautious.

The capability stack reveals why companies with strong product-market fit adapt better. They often already control distribution, integration, and trust. Adding AI features strengthens those moats. Startups that lack these layers may build useful features but struggle to win adoption at scale. Giants that control data and distribution but lack trust or agility may stall.

Closing

Startups struggle in the AI era because they face unfavorable supplier dynamics and lack economies of scale. However, they are still better positioned than the big companies. Giants stall because inertia and slow processes prevent them from moving at the speed AI requires. Firms with proven product market fit, embedded in workflows and trusted by customers, adapt faster because they can integrate AI into what already works and expand from there.

For product leaders, the lesson is clear: advantage in AI does not come from raw technology alone. It comes from combining the right structural powers, navigating market forces, and stacking capabilities where your company is already strong.

In the next post, I’ll turn from why AI disruption plays out this way to how product leaders can respond. We’ll look at practical playbooks like the wedge-to-expansion strategy, jobs-to-be-done, and dual transformation that guide teams through this shift.