The AI boom that’s costing more than it’s worth

Cloud-based AI infrastructure is driving unsustainable costs and market distortions

AI companies are spending billions building global data centers to power generative AI systems. These facilities demand an enormous amount of processing power, storage, and energy. Much of this spending is driven by venture capital, even though many of these firms have yet to turn a profit. In the short term, this buildout seems like progress, more capacity, better performance, faster training times. But there’s a growing imbalance forming between ambition and economics.

Chip manufacturers are redirecting production away from standard memory, the kind used in everyday devices such as laptops and smartphones, toward specialized chips for AI servers, like 3D NAND components. This decision is squeezing supply for consumer electronics, forcing prices higher and reducing what consumers get for their money. Companies are selling products with lower specs to cover rising costs. That dynamic doesn’t just hurt consumers; it affects long-term market equilibrium.

Decision-makers need to see this clearly. The current model locks up capital in large-scale infrastructure that may not pay back quickly, or at all. The supply chain distortion it’s causing highlights a deeper inefficiency: too many eggs in one basket before the economics are proven. Moving some of that investment focus toward smaller, localized AI systems could balance the equation, better cost efficiency without stalling innovation.

According to Gartner, PC shipments will drop 10.4% and smartphone shipments 8.4% in 2026, with respective price hikes of 17% and 13%. TrendForce projects memory prices will rise by 75% beyond the 100% increase already seen. These are not temporary fluctuations, they show structural strain across the industry. If we continue investing in centralized infrastructure without balance, the entire ecosystem becomes fragile. What’s needed now is smarter allocation, not just more spending.

The economics of cloud-based AI are fragile and potentially unsustainable

The current economic model driving AI is reaching its breaking point. Many companies are subsidizing operations at a scale that doesn’t make long-term sense. The numbers are clear: for around $200 per month, users can consume $8,000 worth of Anthropic tokens or $14,000 from OpenAI. That imbalance indicates a cost structure out of alignment with real value creation.

Technology critic Ed Zitron has described the situation as “AI economics already broken,” suggesting that what we’re seeing could be a classic investment bubble. OpenAI, Anthropic, and even Meta are already considering price cuts and tighter limits on internal usage. One Cisco executive admitted publicly that token costs are higher than the business value they currently produce. The Times also reported two large global banks spending $1 billion on AI experiments that yielded almost no measurable returns. These examples point to a financial system under stress, fueled more by excitement than sustainable economics.

For C-suite leaders, this is a warning sign. The industry needs to evolve past its early growth-at-all-costs mindset. It’s time to measure progress by return on value. Subsidized use cases may create impressive demos but don’t always translate to meaningful enterprise adoption. The next wave will depend on making AI viable without subsidizing it into existence.

Leaders should think about resilience. Technology moves fast, and cost structures that look steep today could fall tomorrow, but that transition period can destroy balance sheets if strategy doesn’t adapt. AI is going to reshape business, but the companies that win will be those that build with discipline.

On-device (Edge) AI may render massive data center investments obsolete

The direction of AI is shifting from centralized servers to devices that process information locally. This is called edge AI, where machine learning happens directly on the phone, laptop, or wearable instead of a remote data center. The implications are enormous. It means lower operational costs, reduced latency, more privacy, and less dependence on massive server infrastructure.

Apple is leading this direction. Working with Google Gemini, it distilled large AI models into efficient versions that run directly on its hardware. The result is a more capable Siri, an assistant that can analyze emails, messages, photos, and even pull relevant data from the web, while keeping most of the processing on-device. This work fits with Apple’s long-standing focus on security and privacy. The company has also built a developer ecosystem around its Foundation Models framework and introduced MLX Distributed, which allows different Apple devices to collaborate in running AI tasks. Its acquisition of Liquid AI further strengthens its commitment to this vision.

For executives, this transition matters because edge AI distributes computational load across millions of devices. That drastically reshapes cost dynamics, reducing the need for huge capital investment in cloud infrastructure. Energy use, data management, and compliance risks are also minimized since user data stays local. Companies relying exclusively on cloud-based AI may find their infrastructure outdated faster than anticipated, an issue that can erode returns long before depreciation schedules end.

Tim Cook, CEO of Apple, has already warned of rising component prices, driven by heightened demand for high-value memory parts. But Apple’s shift toward on-device AI shows that a more efficient path exists. It’s not about abandoning the cloud, it’s about recognizing that the edge will handle an increasing share of real-world AI work. The companies that see this early will allocate capital better and build more sustainable business models for the long run.

The AI hardware rush is contributing to consumer price inflation and degraded product value

The industry’s heavy focus on AI servers is reshaping the electronics market in a costly way. Memory suppliers are prioritizing AI-related components, creating shortages of standard parts for consumer devices. That shortage is driving manufacturers to raise prices or quietly reduce performance specifications to keep products affordable, a practice often referred to as “shrinkflation.”

Major players have already taken action. Sony increased the PlayStation 5 price by $100, while Microsoft, Nintendo, and Samsung have all raised device prices across their product lines. Even Apple, known for tight cost control, anticipates an increase of $100 to $150 in the price of the upcoming iPhone 18 Pro. These shifts directly tie back to supply chain constraints. As production capacity remains tight, suppliers can charge more, and few are willing to expand output because of uncertain AI demand longevity.

Executives should view this trend as both a risk and a signal. Rising production costs can hurt consumer sentiment, compress margins, and limit innovation cycles. When essential components grow scarce, manufacturers often must make trade-offs that reduce differentiation and product quality. For companies leading global supply chains, mitigating these effects means securing direct supplier relationships, diversifying production, or investing upstream in memory technology development.

Tim Cook, CEO of Apple, has clearly indicated that upcoming price increases are linked to the demand surge for AI components. Gartner’s projections support this: PC shipments are expected to drop by 10.4% and smartphones by 8.4% in 2026, with average prices increasing by 17% and 13%. The memory component of this story is critical. As AI becomes more integrated into every part of the economy, the tech sector must address supply imbalances or risk a prolonged period of elevated prices and slower growth.

For business leaders, the key takeaway is control, control over cost structure, component strategy, and technological direction. The companies that succeed in this environment will be those that manage scarcity through innovation and forward procurement.

The shift to edge AI poses major consequences for investors in cloud-based infrastructure

AI’s evolution toward on-device computing is moving faster than many investors anticipated. As edge AI becomes more capable, the need for massive cloud infrastructure could decline sharply. Companies have already poured billions into constructing data centers designed for large language models and generative AI workloads. These investments made sense when all major AI computation had to happen on centralized servers. But as local models outperform expectations, the premise behind these infrastructure expenditures is weakening.

When AI workloads move to devices, the economics of large-scale data centers become less compelling. Maintenance, energy consumption, and hardware upgrade cycles are high, and the return horizon keeps stretching. With new generations of processors appearing frequently, even the latest data center servers, such as those built around Nvidia’s H200, will need upgrades sooner than expected. If utilization drops before investors realize their returns, those assets risk losing value quickly. For organizations sitting on extensive cloud infrastructure, this could compress revenue while operational costs remain high.

For executives and investors, the message is straightforward. The future of AI is blending cloud capability with distributed intelligence across devices. Companies that continue to allocate capital primarily toward heavy, centralized infrastructure could find themselves overexposed to technologies approaching maturity. The focus now should be on adaptable systems that bridge cloud and edge AI, maintaining scalability without locking resources into infrastructure that may soon see lower demand.

This transition also creates opportunities. Firms developing hardware optimized for local AI or tools for decentralized computing stand to benefit. Investors should monitor this trend closely and evaluate portfolios accordingly. The winners will be those who anticipate and align with the shift.

Overall, the shift to edge AI is not a distant concept, it’s underway. Those holding on to cloud-centric strategies without diversification risk significant asset devaluation. For decision-makers, the task now is to align digital transformation strategies with where computation is heading, ensuring agility in both technology adoption and capital deployment.

Key takeaways for decision-makers

• Cloud AI investment is driving market imbalance: Overinvestment in cloud infrastructure is straining supply chains and inflating device prices. Leaders should reassess capital allocation toward scalable, sustainable AI models that don’t distort component markets.
• AI’s economic model is unsustainable under current conditions: The cost-to-value gap in cloud-based AI services is widening. Executives should evaluate ROI rigorously and pivot toward solutions that generate measurable operational or financial returns.
• Edge AI is emerging as a smarter, cost-efficient direction: On-device AI reduces dependency on data centers, cuts costs, and enhances privacy. Decision-makers should start integrating edge capabilities into product and infrastructure strategies to improve scalability and resilience.
• Hardware shortages are raising consumer prices and reducing value: AI-driven demand for advanced memory is inflating electronics costs while decreasing performance quality. Leaders should diversify suppliers, secure critical components early, and invest in alternative materials or production capacity.
• The shift from cloud to edge computing carries major investment risks: Existing data center assets may lose value as AI computation decentralizes. Executives and investors should rebalance portfolios toward adaptable, hybrid AI infrastructure that aligns with long-term efficiency and flexibility.