Money, talent, and attention are migrating toward AI at a pace few sectors have ever experienced. Boardrooms that once debated logistics upgrades or international expansion now weigh foundation model partnerships and data pipelines. Venture funds that used to back logistics, healthcare devices, and consumer brands are suddenly “AI-first.” The result is more than a technological shift: it is a reallocation of the basic inputs of growth. That is why this topic cannot be treated as another “AI opportunity” explainer. The real question is harder and more uncomfortable: as AI absorbs more of the world’s scarce growth resources, what does it do to everyone else’s cost of growth?
At the center of this article is a tension that serious investors and operators can no longer ignore. On one side is the argument that AI is a general-purpose technology, so heavy investment today will lift all boats tomorrow. On the other is a quieter but growing concern: that the concentration of capital, talent, and compute into AI is raising the hurdle rates, wage levels, and entry barriers for businesses that are not, and do not need to be, AI-native. Put bluntly: is the AI boom making it more expensive to grow if you are in “boring” but essential sectors like manufacturing, logistics, or non-digital services?
The governing lens we will return to is simple but demanding: the cost of capital and resource allocation efficiency across sectors, expressed through two practical questions. First, how hard and expensive is it becoming to access the funds, people, and infrastructure you need to grow if you are not in AI? Second, how do changing entry barriers reshape the competitive landscape and long-term returns? The analytical claim that follows is explicit: the AI investment boom is increasing the cost of growth for non-AI sectors by reallocating capital, talent, and infrastructure, thereby raising market entry barriers and intensifying competition for limited resources. The argument of this essay is that, given that reality, investors and business leaders should not chase AI at any cost, but instead adopt a consciously balanced portfolio posture that keeps AI in view without starving the rest of the economy.
AI capital concentration and sector crowding
The first and most visible vector is capital. AI is not just attracting investment; it is reorganizing capital markets. Large funds are carving out dedicated AI vehicles, corporate treasuries are redirecting growth budgets into AI experiments, and lenders are upgrading AI-linked projects in their internal priority lists. This is not neutral: when a given pool of money is finite, sending more of it to one domain raises the effective cost of capital for other domains. You see this in term sheets: non-AI companies being pushed to accept higher return expectations, tighter covenants, or slower funding timelines because “we’re reserving capital for AI deals this quarter.” The context is not just enthusiasm—it is a structural repricing of what non-AI growth must earn to compete for capital.
Consider a regional private equity firm that historically backed industrial services, specialty manufacturing, and healthcare practices. As AI valuations soar, the partners launch an AI-focused fund and reserve their best internal talent to source those deals. Suddenly, their industrial prospects find that equity checks are smaller, diligence cycles are longer, and post-deal support is thinner. The AI fund might well earn outsized returns, but for the industrial targets, the practical outcome is higher effective cost of capital and slower growth. The same dynamic this firm experiences is playing out across many institutional allocators: AI is not just a new lane; it is cannibalizing attention and risk capacity from existing ones. In terms of our governing metric, the hurdle rate for non-AI sectors is being ratcheted up, not by their fundamentals, but by capital’s appetite for AI exposure.
The crowding effect also alters the “acceptable story” for receiving funds. Ten years ago, a well-run logistics company with predictable cash flow and a clear expansion path could command healthy multiples because growth plus stability was enough. Today, in many investment committees, a non-AI business is subtly asked to attach an AI tail just to match the expected upside of AI-native peers. If it refuses, it may see discounting on its valuation or struggle to compete for capital against more speculative but AI-branded ventures. This is a quiet but real increase in the cost of growth: not only is capital more expensive, but the narrative requirements have risen too. A logistics operator that would once clear the cost-of-capital bar now fails it unless it promises AI upside, even if the core business economics are unchanged.
From the perspective of resource allocation efficiency, this is where the tension bites. If the marginal dollar invested in AI genuinely produces superior risk-adjusted returns, capital concentration makes sense. But if a portion of that capital is chasing “AI exposure” without disciplined ROI scrutiny, then the system is reallocating from steady, positive-ROI projects in non-AI sectors to uncertain ones in AI. The governing metric—sector-level ROI relative to risk—can deteriorate even as headline AI valuations climb. The causal link is straightforward: every dollar pulled from a profitable, underfunded project in a non-AI sector is a dollar that must be justified by superior risk-adjusted returns in AI. If that superiority is assumed rather than proven, the opportunity cost quietly accumulates. That is why simply tracking AI deal volume misses the point; the sharper question is how often solid non-AI projects are being starved despite strong fundamentals—and how often that starvation is driven by capital fashion rather than rational optimization.
AI talent markets and wage inflation
Capital is only the first bottleneck. The second is talent, especially in technical, product, and data roles. AI has created a high-intensity bidding war for a relatively small pool of specialized skills: machine learning engineers, data scientists, distributed systems architects, and experienced engineering managers capable of leading AI-heavy teams. As AI firms and big tech platforms escalate compensation, they drag up wages for adjacent roles and alter the entire labor market structure—often in ways that directly raise costs for non-AI firms. The issue is not simply higher salaries; it is a repricing of what it takes, in human capital terms, to be a competitive modern business.
Picture a mid-size logistics company trying to modernize its routing software and analytics. Its needs are modest: a couple of solid data engineers and one senior architect to tie systems together. Several years ago, it could hire that team at upper-mid-market salaries. Today, the same candidates receive offers from AI-native companies and AI-focused labs that are dramatically higher, plus equity with lottery-ticket potential. The logistics firm either pays closer to AI-level compensation (raising its cost base and creating internal pay compression), settles for less experienced talent (slowing execution and degrading quality), or postpones the upgrade (delaying productivity gains). Every choice translates into a higher cost of growth or a slower growth path. Through our lens, its internal ROI threshold for digitization projects has shifted: more spend is required to get the same uplift, and some otherwise attractive projects no longer clear the bar.
There is also a subtler brain drain of entrepreneurial and managerial talent. Ambitious product leaders, founders, and investors are pulled into AI ventures not just by money but by prestige and the sense of working on a frontier. That is rational at the individual level, but at the system level it deprives non-AI sectors of exactly the people who translate capital into operational improvements. A manufacturing or healthcare operator that once attracted top graduates for its complex, meaningful problems now competes against AI labs promising world-changing models and outsized equity upside. Over time, that degrades non-AI sectors’ ability to execute ambitious transformations even when they secure funding. The consequence is a compounding effect: higher costs to hire scarce talent and weaker capacity to turn that talent into realized productivity.
From an efficiency standpoint, this raises a hard question that goes beyond anecdotes: are we allocating our scarcest human capital—the people who convert resources into compounding productivity—into the domains where their marginal impact on real output is highest, or into the domains where narrative and equity upside are most compelling? When viewed through our governing lens, the issue is not that AI talent is “overpaid,” but that its wage signal distorts the broader labor market. Non-AI firms face a higher marginal cost of crucial hires, pushing their required ROI thresholds up. Projects that would have cleared the hurdle at last decade’s wage levels are now uneconomic, even if their fundamental value creation remains strong. The system, in effect, raises the “price of admission” for meaningful operational upgrades outside AI-heavy domains, and with it the cost of growth in those sectors.
AI compute bottlenecks and infrastructure costs
The third constraint is compute and related infrastructure: GPUs and accelerators, cloud costs, data pipelines, and the specialized tooling needed to run modern AI at scale. What makes AI different from prior software waves is its voracious appetite for high-end hardware and dense data processing. Capacity is not infinite; the same clusters and chips that train large models also power analytics, simulations, and digital operations across many sectors. As AI demand explodes, non-AI workloads can find themselves paying the price—literally. The relevant context is that digital infrastructure has become a shared input, and AI’s consumption of it is reshaping its pricing and availability.
Consider a large retailer that wants to roll out more granular demand forecasting, customer segmentation, and supply chain optimization. These are not speculative science projects; they directly increase gross margin and reduce working capital. Yet their workloads often rely on the same GPU instances and high-performance cloud services that AI model training consumes. If cloud providers prioritize AI customers willing to sign massive, multi-year commitments, the retailer sees higher spot prices, more frequent capacity constraints, or pressure to commit to long-term minimums that were previously unnecessary. In effect, AI tenants are bidding up the rent on shared infrastructure, raising the cost of digital growth for everyone. A forecasting system that once produced an attractive payback period now faces higher infrastructure inputs, weakening the investment case or stretching the time to breakeven.
There is a physical angle as well. Data centers, power, and cooling have become strategic bottlenecks. When hyperscalers and AI labs sign up entire new campuses to power model training, they may crowd out more mundane but essential needs: regional data centers for industrial IoT, latency-sensitive workloads for logistics or finance, and localized processing for regulated sectors. Regulators and utilities, facing pressure to allocate limited energy and land, may skew in favor of AI because it promises higher economic output per megawatt—on paper. But for a non-AI manufacturer that needs edge compute capacity near a port or factory, this can translate into longer timelines and higher connection fees. The causal chain is clear: AI demand induces infrastructure build-out skewed toward its own needs, and the residual capacity—and pricing—for other digital workloads adjust upward, raising their cost of growth.
Returning to our metric of resource allocation efficiency, compute is a test case in how AI can distort price signals. If GPU clusters flow to the highest immediate bidder, AI projects with speculative revenue potential can outcompete non-AI projects with clearer, but less hyped, ROI. The short-term effect is straightforward: non-AI firms pay more or wait longer for the same infrastructure. The long-term effect is more structural: firms that cannot pre-commit to large AI-style contracts may be structurally disadvantaged in access to the digital infrastructure that underpins modern competitiveness, regardless of whether they ever deploy a model themselves. In aggregate, sector-level cost of growth rises, and market entry barriers for infrastructure-dependent newcomers harden—not because their ideas are weaker, but because the shared input market has been tilted.
Rival narratives and productivity spillovers
So far, the picture might sound like an indictment: AI is hoarding capital, talent, and compute, leaving others to fight over scraps. But the debate is not that simple. There are three serious counter-arguments that deserve to be engaged directly, each offering a competing logic for how this story might end—and, crucially, each resting on different expectations about our governing metric of cost of capital and resource allocation efficiency.
The first is the “AI bubble” thesis: that current capital allocation is mispriced and will inevitably correct. In this view, runaway valuations and resource hoarding will be followed by a washout as many AI ventures fail to generate sustainable margins. When that happens, resources—capital, talent, even compute—will be reallocated back into more traditional sectors at more reasonable prices. Historically, bubbles in one sector have often led to infrastructure and human capital that, after the bust, become cheap inputs for everyone else. Investors backing non-AI firms might then benefit from a buyer’s market in talent and infrastructure, temporarily lowering their cost of growth below historical averages and boosting ROI for those who remained diversified.
The second is the “general-purpose technology” argument. Here the claim is that AI is akin to electrification or the internet: a foundational capability that, once mature, will permeate and improve nearly all sectors. Short-term resource concentration is thus not misallocation but necessary front-loading of experimentation and learning. Non-AI sectors, on this view, will eventually receive AI-driven productivity tools, process automation, and decision-support systems that more than offset the temporary spike in their cost of growth. The higher hurdle rates they face now are compensated by larger productivity gains later; over a full cycle, ROI and resource allocation efficiency improve rather than deteriorate. The causal link runs through spillovers: early concentration enables breakthroughs that later diffuse, lowering everyone’s long-run cost of capital relative to outputs.
The third is the “domain adaptation” claim. Instead of relying on AI spillovers, some argue that non-AI sectors will innovate along different axes—materials, hardware, processes, business models—that do not directly compete with AI for capital and talent. A construction firm might reinvent modular building methods; an agriculture company might pioneer regenerative practices; a logistics company might refine network design. In each case, growth paths may emerge that are relatively insulated from AI, allowing these industries to grow on their own terms even as AI dominates digital discourse. Under this logic, the rising cost of AI-specific inputs matters less because these sectors earn their returns from different, less contested resources, keeping their cost of growth in line with their own innovation cycles.
Each counter-argument has conditions under which it could outperform the central concern of rising growth costs. If AI proves significantly overstated and fails to deliver monetization, the bubble thesis looks prescient; cost of growth in non-AI sectors might fall as capital and talent “return home,” and resource allocation efficiency improves simply through reversion. If AI’s productivity gains turn out to be dramatic and widely diffused, the general-purpose narrative will vindicate today’s capital concentration as farsighted; sector-level ROI will rise enough to justify today’s crowding because downstream productivity drops will offset the earlier resource premiums. If non-AI sectors discover high-return innovation lanes that do not require AI-shaped resources, domain adaptation can sidestep much of the resource crowding, allowing those sectors to maintain or even lower their cost of growth despite AI mania.
The problem is that none of these outcomes is guaranteed, and all operate on longer time horizons than practical capital cycles. For allocators and operators, this uncertainty reinforces the need to keep returning to the governing lens: What is the current and near-term cost of essential inputs for your sector relative to likely returns, given today’s observable dynamics? Betting solely on eventual spillovers is not a strategy; it is faith. At the same time, assuming AI is pure misallocation ignores genuine breakthroughs already visible in certain tasks and workflows. The core tension thus remains live: AI-driven capital concentration versus the broader market’s need for diversified, cost-efficient investment today. A credible strategy has to be built around that tension, not resolved by storytelling about a distant equilibrium.
Entry barriers, competition, and non-AI growth
Where these resource shifts become structurally important is in entry barriers. Even if incumbents can absorb higher costs of capital, talent, and compute, new entrants often cannot. The AI investment boom may thus be hardening market structures in non-AI sectors, not because incumbents are superior, but because the price of contesting them has risen. This is where the cost-of-growth lens connects directly to the competitive landscape and long-run sector dynamism.
Imagine a founder with deep experience in regional logistics who wants to launch a modern, tech-enabled carrier focused on under-served routes. Ten years ago, their main obstacles would have been fleet financing, route access, and recruiting reliable drivers. Today, they must also budget for a competent data team, cloud infrastructure that integrates with large retailers, compliance software, and a baseline digital customer experience that meets heightened expectations shaped by AI-powered incumbents. Each of these line items has become more expensive precisely because of the broader AI pull on resources. The net effect: the break-even scale and capital requirement for entry have climbed, and with them, the minimum acceptable ROI that investors will demand to justify that entry.
This dynamic is not confined to tech-adjacent fields. In healthcare, education, and traditional B2B services, customers raised on AI-powered tools increasingly expect personalization, predictive support, and sophisticated analytics baked into offerings. A new non-AI-native firm is thus forced either to integrate AI capabilities (competing in distorted input markets) or to position itself as deliberately “low-tech” and accept a smaller, often more price-sensitive niche. Entry without AI-influenced capabilities becomes a harder commercial proposition, which translates into higher required capital and talent—back to our core metric of the cost of growth. The “minimum viable” bar in many markets has shifted upward under the pressure of AI-shaped expectations, pushing viable entry further out of reach for under-capitalized teams.
From a competitive landscape perspective, this can entrench incumbents in sectors that historically saw frequent disruption. Large firms can spread the higher cost of AI-adjacent inputs across a broad revenue base, negotiate preferential terms with cloud and chip providers, and recruit AI talent by promising scale and proprietary data. Smaller challengers, in contrast, face those costs upfront with no guarantee of recoupment. The end state may be fewer entrants, slower turnover, and a tilt toward consolidation driven less by fundamental efficiency and more by the side effects of AI-heavy input markets. In terms of resource allocation efficiency, the system risks locking capital into incumbents not because they are the best users of it, but because newcomers cannot clear the new, AI-inflated entry thresholds.
There is a paradox here: AI is often marketed as a democratizing force that lets small firms punch above their weight through automation and smart tools. That can be true for off-the-shelf AI services that lower certain operational costs. But when we step back to our governing lens—sector-wide resource allocation and entry barriers—the picture is more mixed. AI may lower some variable costs while raising fixed costs of competing at a modern standard. For investors and policymakers who care about competitive dynamism rather than just frontier performance, this trade-off should be front and center. A market that becomes more efficient at the top while less contestable at the bottom may show strong margins but weaker long-term innovation and more fragile growth.
Portfolio diversification and decision posture
All of this leads to a strategic question for investors and business leaders: how should you position capital and operating plans in the face of an AI boom that simultaneously promises transformational upside and raises the cost of growth elsewhere? The argument here is not to avoid AI; that would ignore both the opportunities in AI itself and the real risk of being left behind where AI becomes embedded infrastructure. Rather, the case is for a disciplined, balanced posture that treats AI-heavy bets as one part of a diversified portfolio, not the entire thesis.
For investors, a practical lens is to compare the risk-adjusted ROI of AI-centric investments with that of high-quality non-AI opportunities, explicitly factoring in rising input costs. A simple rule-of-thumb could be framed as:
effective ROI = expected project ROI − (capital premium + talent premium + infrastructure premium relative to sector baseline).
When you apply that consistently, you may find that some non-AI plays with temporarily elevated wage or compute costs still clear your threshold more comfortably than AI ventures that depend on unproven monetization. Allocating a meaningful share of capital to those undervalued, cash-generative opportunities can both stabilize portfolio returns and leave you better positioned if AI valuations compress or input markets normalize, because your capital is not entirely tied to the most resource-intensive corner of the economy.
For operating businesses, the analogous question is resource allocation inside the firm. How much of your growth budget, senior leadership attention, and hiring capacity should be committed to AI initiatives versus core operational improvements that may have clearer payback but feel less exciting? One disciplined approach is to subject AI projects to the same hurdle rates and time-to-payback expectations as other capital expenditures, rather than assuming a “strategic” exemption. If an AI initiative cannot justify its resource draw under the same governing metric—its contribution to lowering your effective cost of growth and raising ROI—it should not leapfrog projects that can. That discipline is the practical expression of a balanced stance, and it directly resists the narrative pressure to treat AI as exempt from normal capital discipline.
A balanced posture also improves resilience against the unknowns we identified: the pace of AI maturity, regulatory responses, and the possibility of breakthroughs in non-AI domains. If AI regulatory regimes tighten or public sentiment shifts, AI-heavy valuations could reset rapidly, but the resource distortions (wage structures, data center build-outs) could linger. Firms whose fortunes are entirely tied to AI narratives would be exposed, while those with diversified exposure and strong non-AI engines could redeploy freed-up talent and infrastructure into more conventional but reliable growth paths. From a portfolio perspective, diversification is not about hedging against AI as a technology; it is about ensuring that the rising cost of growth in one domain does not drag your entire strategy off course.
The condition that could plausibly change this judgment is a clear, sustained demonstration that AI-driven productivity is so strong and so broadly diffused that resource constraints ease on their own. If AI tools massively amplify non-specialist talent, reduce the need for scarce experts, and optimize infrastructure utilization, the current crowding effects might reverse; capital might become cheaper as productivity rises, wages might decouple from niche skill scarcity, and compute bottlenecks might ease through efficiency gains. In that world, a more aggressive AI-biased allocation could be warranted because the cost-of-growth penalty on non-AI sectors would diminish, and resource allocation efficiency would improve across the board.
But we are not there yet, and strategy cannot be based on hoped-for equilibria. For now, the weight of evidence points to an AI investment boom that is raising the cost of growth for many others by pushing up the price of capital, talent, and infrastructure and by hardening market entry barriers. Against that backdrop, the seemingly “neutral” stance—piling into AI like everyone else and assuming spillovers will sort things out—is not neutral at all; it is a choice to accept higher systemic risk and weaker diversification.
The more defensible judgment is to adopt a deliberate balance: commit capital and attention to AI where the risk-adjusted ROI justifies the resource premium, while preserving substantial, patient investment in non-AI sectors that are currently underpriced relative to their long-term value creation. That means repeatedly asking, for every significant allocation: does this deployment of scarce resources genuinely improve long-term, risk-adjusted returns once we account for the rising cost of growth, or is it simply following the loudest story in the room? The answer to that question—not enthusiasm about AI in the abstract—should anchor how serious investors and operators act in this phase of the AI era.