Daiwa just dropped a target price cut on Tencent, slashing it to 670 HKD while slapping a “Buy” rating on the stock. The headline says “lower target, higher conviction.” Sounds like a contradiction? It’s a trap. The real story is the capital expenditure revision: 2026 AI CapEx jacked from 108 billion RMB to 181 billion RMB. That’s a 67% upward revision. In DeFi terms, think of it as a protocol deciding to lock 67% more TVL into a single, illiquid vault with a 2-year lock-up period. The underlying asset? GPUs. The yield? Unproven. The risk? Structural.
Let me be clear: I don’t trade stocks. But I read balance sheets like smart contracts. And this balance sheet is telling me Tencent is undergoing a paradigm shift from a cash-flow farming machine (advertising, gaming, payments) into a capital-intensive infrastructure provider. That’s the same shift we saw in crypto when everyone moved from pure speculation to yield farming — and then got wrecked when the underlying asset (LUNA) collapsed. The difference? Tencent has real cash flows. But the question is: can they sustain this burn rate without triggering a liquidity crisis?
Here’s my framework. Tencent’s current revenue streams — gaming, social ads, cloud — are like a high-yield stablecoin pool. They generate steady, predictable yield (30-40% margins). The AI CapEx is a leveraged bet on a new primitive: inference-as-a-service. They’re taking the stablecoin yield and pouring it into a volatile, high-risk vault (GPUs, data centers, model training). The depreciation on those GPUs? That’s the impermanent loss. If AI demand doesn’t materialize by H2 2026, the depreciation will eat into the base yield. This is the same math I used when analyzing EigenLayer restaking — you need to account for slashing risk (obsolescence, competition, regulatory) before calculating net yield.
Liquidity doesn’t care about your roadmap. Daiwa believes Tencent’s core biz (gaming, ads) will provide sufficient “buffer” to absorb the depreciation hit. I’ve seen this before: in 2020, Compound’s oracle latency looked fine in backtests, but under real gas wars, the buffer evaporated. I spent 72 hours simulating that exploit. Tencent’s buffer is its gaming revenue, which is already decelerating from a high base. If a recession hits (consumer spending drops, ad budgets shrink), the buffer shrinks. Simultaneously, the AI CapEx is fixed. That’s a negative convexity position — your liabilities (depreciation) are linear, but your assets (revenue) are convex to macro headwinds. In DeFi, that’s called a “liquidation cascade waiting to happen.”
Let’s zoom into the CapEx structure. Daiwa mentions “chip supply improvements.” This is code for “Tencent found a way to buy Nvidia H100s or is scaling domestic alternatives like Huawei Ascend.” Either way, the availability of hardware is not the bottleneck — it’s the operational cost of running a massive AI inference network. The real technical debt is software-level: rewriting WeChat, QQ, advertising engines, and gaming servers to pipe through AI models. That’s not a CapEx line item; it’s an OpEx hemorrhage. Based on my audit experience in 2017 with Mantra21’s voting contract, I know that code bloat kills efficiency faster than hardware upgrades. Tencent is essentially issuing a “technical debt bond” that matures in 2027. The comfort level? Low.
Now, the contrarian angle. Everyone says “AI is the future, so invest now.” That’s retail thinking. Smart money knows that timing matters more than conviction. In 2022, when Terra fell, I didn’t panic sell — I analyzed the algorithmic stability module and realized the feedback loop was broken. I shorted PAXG and BTC perps, preserving 80% of capital. Tencent’s AI bet has a similar structural flaw: the monetization timeline depends on inference demand exploding in H2 2026. That’s a specific, fragile assumption. If the killer AI app doesn’t arrive (or arrives earlier to a competitor like ByteDance), Tencent’s ROI craters. The counterparty? Retail investors who think they’re buying a tech giant but are actually buying a venture capital vehicle with a 2-year lock-up.
I don’t fight the Fed, but I front-run the machine. In DeFi, we assess yield opportunities by stress-testing the protocol’s tokenomics under multiple scenarios. Tencent’s scenario matrix: - Scenario A (bull): AI demand explodes, cloud revenue jumps 30% CAGR, depreciation is covered, margins stabilize. Tencent wins. HKD 670 looks cheap. - Scenario B (base): AI demand grows linearly, depreciation drags margins down 100-200 bps, stock trades sideways. Target price 670 is fair. - Scenario C (bear): Recession hits, gaming/ads revenue drops 10%, AI CapEx can’t be recouped, earnings fall 20-30%. Stock drops to 450 HKD.
The market is pricing Scenario A with a 50% probability. I’d assign 30%. Why? Because I’ve seen this movie in DeFi: protocols that overspend on infrastructure (e.g., L2 sequencers buying centralized servers) and then fail to achieve decentralization because the capex doesn’t translate to user adoption. The ledger doesn’t lie — Tencent’s balance sheet shows a massive increase in PP&E relative to revenue growth. That’s a red flag for operational leverage.
Let’s talk about the timing. Daiwa expects AI monetization to begin H2 2026. That’s 18 months from now. In crypto terms, that’s an eternity. The average DeFi yield strategy pivots every 3 months. Tencent’s capital is locked into GPUs with a 4-year depreciation schedule. If the AI narrative shifts (e.g., a new architecture like Mamba replaces transformers), those GPUs become stranded assets. This is exactly what happened to mining farms after ETH transitioned to Proof-of-Stake — billions in GPUs became worthless. Tencent’s risk? They’re building a mining farm for a consensus mechanism that hasn’t been finalized.
The key signal to track is not the CapEx number. It’s the utilization rate of their AI inference clusters. If they can achieve >70% utilization with paying clients (e.g., enterprise API calls), then the unit economics work. If utilization stays below 40%, the depreciation becomes a deadweight. I’ve built tools to audit autonomous wallet behavior — I’d apply the same logic here: monitor on-chain data (if Tencent publishes cloud revenue breakdown). Unfortunately, they don’t. So we’re flying blind.
Panic sells, patience profits, code protects — but in this case, there’s no code. Just a PowerPoint of a future AI ecosystem. Tencent’s stock is built on faith in management’s ability to execute. And after auditing three major DeFi protocols that promised revolutionary tech and delivered buggy contracts, my faith is calibrated. I’m not saying Tencent will fail. I’m saying the risk-reward is asymmetric to the downside given the macro environment.
So what’s the actionable takeaway? If you’re a trader, treat this as a binary option with a 2-year expiry. The price of the option is the current premium (current stock price vs. Daiwa target). The strike is H2 2026 monetization success. I’d rather buy protection (out-of-money puts) than chase the upside. In my own portfolio, I’m allocating to small-cap AI infrastructure plays that can pivot faster. Tencent is the whale — impressive, but slow. And in a volatile market, speed beats size.
I don’t care about the target price. I care about the capital efficiency of the bet. Tencent is deploying $181B of its shareholders’ capital on a yield strategy with an uncertain payoff. That’s a huge allocation to one asset class. In DeFi, that’s called a concentration risk. And concentration risk, when hit by tail events, wipes out your entire portfolio. History says don’t bet the farm on a single thesis.
Liquidity doesn’t care about your roadmap. But it will care when the margin calls start.
— Abigail Thomas, DeFi Yield Strategist, Kuala Lumpur
