“Hotelling’s law states that competing businesses tend to make their products, services, or locations as similar as possible to capture the maximum market share. First proposed by Harold Hotelling in 1929, this ‘principle of minimum differentiation’ explains why rival vendors, brands, or political candidates naturally gravitate toward the centre of a market rather than spreading out to serve distinct areas or niches.” – Hotelling’s law – Economics
The competitive pressure can be understood through a spatial metaphor: firms are not only competing over customers, but also over proximity to the market’s most valuable resources. In a fundraising environment, those resources include investor attention, technical talent, distribution partnerships, benchmark visibility and perceived legitimacy. When capital concentrates around a small group of prominent AI companies, rival ventures may move towards the same institutional and technological centre because proximity improves access, even when it makes their offerings look more alike 1.
Hotelling’s law provides a model for this tendency. In its original form, Harold Hotelling imagined two sellers positioned along a line, with consumers distributed between them and incurring a cost when travelling to buy from either seller. A consumer chooses the option with the lowest combined price and travel cost. Each seller therefore has an incentive to move towards the middle, because doing so makes it harder for the rival to capture customers on the opposite side. The resulting pattern is minimum differentiation: competitors cluster rather than occupy evenly separated positions 2,9.
The mechanism is clearer when location is treated as a proxy for product characteristics. A business may choose a physical site, a political position, a software feature set or a target customer segment. If buyers regard alternatives as close substitutes, moving towards the rival can capture customers who would otherwise face a larger switching or access cost. In a simple two-firm model, the market position of a firm can be represented by x_i, while a consumer’s position is x. The effective cost of purchase is p_i+t|x-x_i|, where p_i is price and t measures sensitivity to distance. The absolute-distance term captures the idea that difference becomes a disadvantage when customers care about convenience or compatibility.
The same logic can apply to investment markets, although the relevant distance is institutional rather than geographic. A start-up that works on a fashionable model architecture, uses familiar performance metrics and recruits from recognised laboratories may be easier for investors to evaluate. Similarity reduces perceived uncertainty and allows a fund to compare opportunities within an established category. It can also generate network effects: investors, employees and commercial partners gather where other high-value participants already gather. The source’s account of AI fundraising therefore fits Hotelling’s intuition, but with capital and credibility functioning as the scarce market space 1.
Why clustering can be rational
Clustering does not necessarily mean that executives believe imitation is strategically ideal. It may be the outcome of defensive positioning. If a competitor occupies a commercially important centre, a distant alternative may reach fewer buyers, attract less talent or appear less compatible with prevailing infrastructure. In a technology market, compatibility with dominant cloud platforms, application programming interfaces, evaluation standards and procurement practices can matter more than nominal differentiation. A firm may then adopt a similar technical or commercial position simply to remain visible and investable.
Hotelling’s result also explains why apparent variety can conceal substantial sameness. Companies may use different branding while converging on similar pricing, feature priorities and distribution channels. Political candidates can move towards the median voter; retailers can open near established shopping destinations; restaurants can gather in recognised dining districts. In each case, the centre supplies dense demand, but competition for that centre can reduce meaningful choice. Hotelling described this as excessive sameness, a result that remains useful for analysing markets where firms compete aggressively yet offer products that appear increasingly interchangeable 2,9.
The model’s limits
The law is not a universal prediction that every competitor will converge. Its original assumptions are restrictive: two firms, a bounded linear market, uniformly distributed consumers, identical products, price-inelastic demand and linear transport costs. Later work showed that the equilibrium can fail to exist under some conditions, while alternative transport-cost assumptions can produce maximum rather than minimum differentiation 2. Price-sensitive consumers, heterogeneous preferences, capacity constraints and strong brand identities can all make distance from a rival profitable rather than dangerous.
Competition also changes when more than two firms enter. Research on spatial competition finds that the central predictions of a two-firm model do not automatically survive in markets with three or more firms. Firms may form local clusters, occupy several differentiated positions or avoid direct proximity when congestion and price rivalry become severe 3,6. Agent-based studies likewise show that outcomes depend on how consumers are distributed, and that the analytical model’s equilibrium conditions may be unlikely to hold in real markets 5. These qualifications turn Hotelling’s law from a rigid rule into a benchmark for identifying the forces that encourage convergence.
From differentiation to strategic distance
The practical lesson is not that differentiation should always be maximised. A business must distinguish between dimensions on which similarity creates access and dimensions on which distinctiveness creates value. An AI company may need to conform to common safety practices, deployment standards and customer workflows, while differentiating through reliability, domain expertise, data rights or operating economics. Convergence around a technical category can lower education costs for buyers, but it also makes a company vulnerable if investors and customers cannot identify a defensible advantage.
There is a further collective risk. When capital flows disproportionately towards a small set of perceived leaders, the market may reward proximity to incumbents instead of independent evidence of product quality. Reports on AI funding describe large late-stage firms absorbing a growing share of available capital, leaving earlier companies to compete more intensely for attention and validation 4. This resembles a positive feedback loop rather than a simple location decision: visibility attracts money, money attracts talent, talent improves visibility, and the resulting concentration makes alternatives appear riskier.
Hotelling’s law still matters because it links individual rationality with potentially undesirable market outcomes. Each firm may have a sensible reason to move towards the centre, yet the collective result can be overcrowding, reduced variety and weaker experimentation. The framework is most useful when applied diagnostically: identify what constitutes distance, determine which resources are concentrated, and ask whether customers benefit from similarity or merely tolerate it. In AI fundraising and other high-attention markets, the central strategic question is therefore not whether firms are converging, but whether the centre reflects genuine demand or an investment ecosystem that has made proximity appear safer than difference.
References
1. Anthropic and the AI cluster effect in fundraising – https://www.ft.com/content/689a2114-57c5-4391-a335-dc97e5716610
2. https://helda.helsinki.fi – https://helda.helsinki.fi/server/api/core/bitstreams/e809df51-49bb-4800-84ef-7e8e91fde9c8/content
3. Hotelling’s Spatial Competition Reconsidered – 2009-02-02 – https://ideas.repec.org/p/tky/fseres/2009cf674.html
4. AI Crowds Out Startups – 2026-03-09 – https://labusinessjournal.com/featured/ai-crowds-out-startups/
5. Agent-Based Spatial Competition – https://www.econstor.eu/bitstream/10419/120490/1/ERSA2012_0156.pdf
6. Spatial competition with interacting agents – 2017-12-31 – https://www.microsimulation.pub/articles/00166
7. Approximation and the Principle of Minimum Differentiation – https://www.ifaamas.org/Proceedings/aamas2019/pdfs/p1949.pdf
8. Hotelling was right with decreasing returns to scale and a … – 2013-02-02 – https://ideas.repec.org/a/spr/anresc/v50y2013i3p953-971.html
9. people.duke.edu · ~dbr1 · researchHotellings-Law-2013-0129-web – Duke University – https://people.duke.edu/~dbr1/research/Hotellings-Law.pdf
10. Hotelling Meets Keynes: Aggregate Adjustment with Spatial Competition and Nominal Rigidity – 2025-03-31 – https://www.nber.org/papers/w33628
11. AI Startup Trends 2026: 6 Funding Shifts for Founders – https://qubit.capital/blog/ai-startup-fundraising-trends
