On August 14, 2025, CITIC Securities International published a research report on Tencent. The headline: Q2 earnings slightly beat expectations. The subtext: a capital expenditure forecast for 2027 raised to 260 billion HKD. That is not a number. That is a signal. And for anyone auditing blockchain projects that claim to be 'AI-powered,' it is a warning etched in depreciation schedules.
I have spent the last two years auditing smart contracts and tokenomics. I have seen teams burn through 10 million USD in six months on 'AI integration' without a single line of production code. Tencent's situation is not identical—it is a mature conglomerate. But the structural dynamics are identical. Capital expenditure growth without corresponding revenue growth is a liability. The code does not lie, only the whitepaper does.
Context: The Four Pillars of Tencent's AI Strategy
Tencent clarified its AI strategy in the earnings call: (1) Large language model training for internal efficiency, (2) AI-powered advertising optimization, (3) AI-driven content creation for gaming and media, and (4) enterprise AI solutions through WeChat ecosystem. These are not speculative. They are defensive. Competitors like ByteDance and Alibaba have already deployed AI in production. Tencent is playing catch-up. The problem is that catch-up is expensive.
CITIC Securities raised its 2026 and 2027 capex forecasts to 215.7 billion and 260 billion HKD, respectively. That is a 42% increase from the previous year. For context, Tencent's core net profit growth in 2026 is expected to be only 2%. The depreciation costs from all that hardware will eat into margins. The report itself lowered core net profit estimates for 2026-2028 by 5% to 9%. This is not a contradiction. It is a mathematical inevitability.
Trust is a variable, verification is a constant. Tencent's financials are auditable. The depreciation line is a verifiable liability. The question is: how many blockchain projects are as transparent? I have read the implementation, not the intent. In my audits, I have seen 'AI layer' projects with zero GPU infrastructure, claiming to train models on decentralized compute. Their capex is zero. Their revenue is zero. But their token valuation is in the billions.
Core: A Systematic Teardown of the Capital Allocation Mismatch
Let me break down the numbers. Tencent's 2027 capex of 260 billion HKD is roughly 33 billion USD. Assume a 5-year straight-line depreciation. That adds 6.6 billion USD in annual depreciation. Tencent's core net profit in 2026 is projected to be around 180 billion HKD based on the report's 2% growth. That means depreciation alone will consume 15% of net profit. That is a drag. The report acknowledges this by lowering estimates.
Now apply this logic to a blockchain project. Suppose a project raises 100 million USD in a token sale, promising to build a decentralized AI training network. They spend 80 million on GPUs and cloud compute. They have no revenue, only token emissions. The depreciation of those GPUs is 16 million per year. Meanwhile, token inflation dilutes holders. The project's 'net profit' is negative, but they hide it behind 'non-GAAP adjusted earnings.' I have seen this script. The code does not lie, only the whitepaper does.
Based on my audit experience, I reviewed a project last year called 'AetherNet,' which claimed to be a Layer-2 for AI inference. They had a partnership with a GPU provider. Their tokenomics showed no capex line. I asked for the smart contract for the GPU rental. They had none. They were using a centralized API. The founders said 'we will decentralize later.' Silence is not agreement, it is data. They raised 30 million. I flagged the risk. Three months later, the GPU provider changed terms, and the project collapsed. The ledger remembers what the founders forget.
Tencent is different. They have real revenue. Their gaming and advertising businesses are growing. The report states that excluding AI investment, operating profit increased 19% year-on-year. That means the core business is strong enough to absorb the capex. But the margin for error is thin. If the AI investments do not yield returns within 2-3 years, the depreciation will become a permanent drag. This is a classic 'capital allocation trap' that every tech company faces. Blockchain projects, however, have no such cushion. They have no advertising revenue. They have no gaming profits. They have only token holders and speculation.
The Contrarian Angle: What the Bulls Get Right
The bulls argue that Tencent's AI investment is a defensive moat. They are right. The AI ecosystem is winner-take-most. Tencent has the data, the distribution, and the regulatory license to operate in China. The report maintains a 'Buy' rating, albeit with a lower target price of 620 HKD from 632. The downgrade is only 2%. That suggests the market is pricing in the long-term value of AI.
But the blind spot is the depreciation timeline. Most analysts treat capex as a one-time investment. It is not. It is a recurring liability. The hardware depreciates, and the software stack becomes obsolete. In crypto, the equivalent is the 'developer burn rate.' I have seen projects hire 50 engineers at $200k each, then run out of funds in 18 months. The bulls say 'they are building for the future.' The reality is that building without a revenue model is a donation, not an investment.
Precision is the only form of respect. Tencent's report provides concrete numbers. It allows me to audit the sustainability. Most blockchain projects do not. They give vague 'roadmaps' and 'token use cases.' I recently read a whitepaper for a project that claimed to 'leverage AI to optimize cross-chain liquidity.' There was no technical specification. No audit trail. No capex estimate. The founding team had a history of rug pulls. The code was a fork of an old Uniswap contract. The market cap was 200 million. The code does not lie, only the whitepaper does.
Takeaway: The Accountability Call
Tencent's capital allocation is auditable. Its financial statements are public. Its depreciation schedule is a line item. Blockchain projects have no such accountability. They can claim 'AI integration' without any verifiable expenditure. The SEC's regulation-by-enforcement is not ignorance of technology—it is deliberately withholding clear rules. But that does not excuse the industry. We need to demand transparency.
In the bear market, only the audited survive. Tencent will survive. It has the cash flow. But the lesson for blockchain is clear: capital expenditure without revenue is a slow death. I am not saying all AI-crypto projects are scams. I am saying you need to verify the capex, the depreciation, and the revenue model. Trust is a variable, verification is a constant. The ledger remembers what the founders forget. Next time you see a project with a fancy AI white paper, ask for the balance sheet. If they cannot produce it, walk away. The code does not lie, but the humans do.