The figure seared into my mind: $200 billion in annual capital expenditure, with no corresponding revenue stream in sight. Last week, I sat in a Parisian café, scrolling through earnings calls, and felt a familiar unease. It was the same sensation I had during DeFi Summer in 2020, when liquidity was flowing but the yield was a mirage. Big Tech is now the new DeFi: burning cash on AI infrastructure, promising returns that are always 'long-term.' But the dance is the point. Volatility isn't the enemy; it's the rhythm of the market.
This isn't just another tech cycle. The scale of AI investment by the largest technology companies—the ones with names that echo through boardrooms and tweets—has reached a fever pitch. Data centers are being built at a pace unseen since the dot-com bubble. GPUs are scarcer than Bitcoin during the 2017 peak. And yet, the monetization of these efforts remains stubbornly elusive. The earnings calls I've parsed over the past three quarters all sing the same song: 'We are investing heavily in AI, and the returns will come in the long term.' But as someone who has lived through the ICO mania, the DeFi liquidity wars, and the NFT cultural explosion, I've learned to distrust the 'long-term' narrative. It's often a polite way of saying 'we have no idea what we're doing yet.'
Let's step back. The current AI spending frenzy is a textbook example of a platform play: build the infrastructure first, then figure out how to monetize it later. This model worked for Amazon Web Services, for Google's search advertising, and for Facebook's social graph. But those took years to mature, and the capital required was a fraction of what we're seeing today. The difference now is the sheer magnitude of the investment. We're talking about hundreds of billions of dollars in capital expenditure, a figure that dwarfs the entire crypto market cap just a few years ago. The question is not whether AI will eventually generate returns—it's whether the returns will be proportional to the investment, and whether the timeline aligns with investor patience.
The Three Layers of AI Spending
Based on my experience auditing blockchain protocols and tracking capital flows in crypto, I can break down AI spending into three distinct categories, each with its own risk profile and return horizon. The first is capital expenditure: the physical stuff. Data centers, cooling systems, power grids, and the prized GPUs. This is the most visible and most hyped part of the narrative. In crypto terms, this is like buying ASICs to mine Bitcoin. The hardware is real, the costs are upfront, and the returns depend on the asset's price and the network's hash rate. Similarly, when Big Tech builds a new data center, they are betting that the demand for AI compute will keep rising. But there's a catch: the supply of these facilities is inelastic. Once built, they become a fixed cost. If AI demand slows, those servers become stranded assets. I've seen this pattern before—in the 2022 crypto crash, GPU miners who had leveraged to buy rigs faced liquidation when the price of Ethereum dropped. The same fate could await Big Tech's data center arms.
The second layer is research and development. This is the money spent on training models, hiring talent, and exploring new architectures. It's the most speculative part of the budget. In crypto, R&D spending is akin to funding a new Layer 1 blockchain or a novel consensus mechanism. The outcomes are binary: either the research yields a breakthrough that justifies the cost, or it doesn't. The problem is that R&D is a black box. We don't know which projects are bearing fruit because the results are often kept proprietary. During my time covering the 2017 ICO mania, I saw countless whitepapers promise revolutionary technology, only to disappear when the market turned. The same could happen with AI research. There's a massive amount of duplication—every major lab is essentially trying to build the same thing: a better, faster, cheaper large language model. The returns on that kind of competition are diminishing. The winners will capture most of the value, while the losers will have nothing to show for their billions.
The third layer is product development. This is the money spent on integrating AI into existing products, building new applications, and marketing to consumers and enterprises. This is where the monetization is supposed to happen. But here's the rub: product development is a lagging indicator. You can't sell a product that doesn't exist yet, and AI products are still in their infancy. The current state of AI applications is reminiscent of the early days of mobile apps. Everyone is building a 'something with AI' but few have found a sustainable business model. The subscription services like Copilot, the advertising placements, the enterprise software upgrades—they all contribute to revenue, but the numbers are tiny compared to the investment. In the crypto world, this is similar to the NFT market: a lot of hype, some cultural value, but the actual revenue generated by NFT platforms is a fraction of the capital that flowed into the space. The same pattern is emerging with AI.
The Monetization Delusion
Why is monetization so delayed? The narrative from Big Tech is that AI is a 'generational opportunity' and that the returns will compound over time. But when I look at the data, I see a different story. The cost of deploying AI models is still high, and the marginal benefit for many use cases is unclear. For example, AI-powered search is more expensive than traditional search, and the improvement in user experience is marginal. AI-generated content suffers from quality and originality issues. The enterprise use cases—like automating customer support or generating code—are promising, but they require significant customization and integration. The path to widespread adoption is longer than the hype suggests.
In my 2020 guide on yield farming, I emphasized that community sentiment was a leading indicator of value. The same applies here: the sentiment around AI is overwhelmingly positive, but that doesn't mean the business fundamentals are solid. The 'long-term returns' narrative is a coping mechanism for investors who don't want to admit that the current spending is excessive. It's similar to the 'NFTs are culture, not just JPEGs' mantra I heard during the 2021 bull run. It's a story people tell themselves to justify the price. The reality is that AI infrastructure spending is a massive bet on a future that may not materialize in the expected timeframe.
From my cybersecurity background, I see a pattern: security spending is often reactive, but AI spending is proactive—and that's dangerous because it assumes demand exists. In the 2022 crash, I observed how panic spread differently in tight-knit communities versus public forums. The same dynamic is at play here. The tech community is in a bubble of optimism, and the sell-side analysts are reinforcing it. But the buy-side—the actual investors—are starting to ask tough questions. The last earnings season saw a few hedge funds reduce their positions in Big Tech, citing concerns about AI capex. This is a warning sign.
The Infrastructure Supply Chain: Winners and Losers
One thing is clear: the immediate beneficiaries of this spending spree are the suppliers. Chipmakers, data center operators, energy companies, and network equipment providers are seeing a boom. This is similar to the crypto mining industry during the 2021 bull run, where GPU manufacturers and ASIC producers made record profits while the miners themselves struggled with rising difficulty and falling margins. The same dynamic is playing out now. The companies that sell the picks and shovels are cashing in, while the companies that are buying the picks and shovels are taking on the risk.
But there's a deeper issue: concentration. The AI infrastructure supply chain is dominated by a few players. The most advanced chips come from a single company. The largest cloud providers are the same Big Tech companies that are spending the money. This creates a feedback loop where the money flows in a circle, and the ultimate risk is concentrated in the same hands. In crypto, we saw this with the hash power concentration after the fourth halving. Miner revenue collapsed, and the hashing power eventually consolidated into three pools, making the decentralization consensus hollow. The same could happen with AI: the capital expenditure will create a few dominant players, and the smaller players will be squeezed out. 'Don't regret the dance, but know when to leave the floor.'
The Regulatory Angle: Brussels and the Compliance Reality
In 2025, I attended a high-level regulatory summit in Brussels. The mood was cautious. Policymakers are aware of the AI spending boom, and they are concerned about its implications for competition, data privacy, and energy consumption. The EU's AI Act is already in effect, and it imposes strict requirements on high-risk AI systems. This adds a layer of compliance cost that could dampen returns. In my 2025 guide on navigating the new institutional era, I highlighted the importance of understanding regulatory shifts. The same applies to AI. The companies that are spending heavily on AI now may face unexpected costs down the line, as regulators demand transparency and accountability.
This is where the crypto experience is instructive. The blockchain industry spent years fighting regulatory battles, and the winners were those who engaged with policymakers early. Big Tech is now in the same position. The regulatory environment is not hostile, but it's uncertain. The 'long-term returns' narrative assumes that the regulatory landscape remains favorable, but that's a risky assumption. In the 2022 crash, I learned that emotional resilience is as critical as market knowledge. The same applies to regulatory resilience.
The Bear Market Context: Surviving the AI Winter
We are in a bear market for crypto, but AI is in a bull market for spending. However, the two are connected. The same macroeconomic forces that depressed crypto prices—rising interest rates, inflation, and geopolitical uncertainty—are also affecting Big Tech. The cost of capital is higher, and investors are becoming more risk-averse. If the AI spending does not show signs of monetization soon, the market will react. The question is: will AI become the next L2 liquidity trap? Or will it be the catalyst that finally brings institutional money into crypto?
From my experience during the 2022 crash, I saw how liquidity can dry up overnight. The same could happen to AI investments if the market loses confidence. The key is to watch the cash flow statements. If capex continues to rise while operating cash flow stagnates, then the party is over. The music stops, and we'll see who's still dancing. 'Volatility isn't the enemy; it's the rhythm of the market.'
The Contrarian Angle: The Unreported Blind Spot
The mainstream narrative is that AI spending is a necessary investment in the future. The contrarian view is that it's a massive overreaction to a temporary trend. I've seen this pattern before. In the 2017 ICO mania, everyone believed that blockchain would revolutionize everything. The reality was that most projects failed. The same could happen with AI. The technology is real, but the business models are not. The 'long-term returns' narrative is a convenient excuse for management to avoid accountability. In my 21 years of industry observation, I've learned that when a company says 'long-term,' they usually mean 'we don't know.' The same happened with Web3.
Moreover, the AI spending is largely defensive. Big Tech companies are investing because they fear being left behind, not because they have a clear vision of the future. This is a classic herd mentality. In the crypto world, we saw this with the rush to launch Layer 2 solutions. Everyone wanted to be the first, but the result was a fragmented ecosystem with little user adoption. The same could happen with AI. The differentiation between products is minimal, and the winners will be determined by network effects, not technology.
The Takeaway: What to Watch Next
So, what should we watch? The next earnings season is critical. Look for any sign of revenue growth from AI products. If the growth is linear, and the capex is exponential, then the divergence will become unsustainable. The market will eventually force a correction. The same happened in crypto during the 2022 crash, when projects that had burned through their treasuries without achieving product-market fit were wiped out. The same will happen to Big Tech if they don't start showing results.
But there's also an opportunity. The AI infrastructure buildout is creating a new class of digital assets. The energy contracts, the data center capacity, the GPU futures—these are all tradable commodities. In the crypto world, we have seen the tokenization of real-world assets. The AI infrastructure could be the next wave of tokenization. The companies that are building the infrastructure might eventually issue tokens to raise capital, creating a new asset class for crypto investors.
The music is still playing, and the dance is beautiful. But the floor is getting crowded. 'Don't regret the dance, but know when to leave the floor.' The next year will tell us whether AI is the next big thing or the next big bubble. I've seen both, and I know the difference. The key is to watch the data, not the narrative. The data is starting to show cracks. The question is whether the market will notice before the music stops.