Tencent Holdings will see revenue grow by 14%

While AI is useful, it cannot replace our critical thinking. I was wrong to accept Copilot's summary of Q1 2026 as valid. The market has continued to decline, and in investigating the reasons, I've found my mistake. I now see that my three-year valuation of Tencent was overvalued and that I still lack the information to estimate its five- or ten-year valuation.

🧠 1. Similar cases: Amazon and Meta

All three situations share the same structural shock:

👉 A mature, cash-generating company 👉 Decides to aggressively reinvest (capex / opex) 👉 The market penalizes it due to lower near-term FCF and margins

Initial outcome:

• 📉 stock decline
• 📉 multiple compression
• ❗ increase in perceived risk

👉 Then one of two paths emerges:

• ✅ successful execution → strong re-rating
• ❌ capital misallocation → value trap

📉 2. META (2022–2023) — the closest parallel to Tencent

📊 What Meta did

• Massive investment in:
  • Metaverse (Reality Labs)
  • AI
• Capex + opex surged sharply

👉 Result:

• Earnings collapsed in relative terms
• FCF dropped significantly

📉 Market reaction

• Stock fell ~70% in 2022
• P/E compressed to very low levels
• Narrative:“Zuckerberg is destroying shareholder value”

🧾 Official management message (very similar to Tencent)

• “We are investing for the future”
• “Returns will come in the long term”
• “The core business remains strong”

🔄 Inflection point

When Meta:

• Reduced spending (restored discipline)
• Improved efficiency
• Maintained core growth

👉 Result:

• Strong re-rating (2023–2024)
• Multiple expansion + margin recovery

🧠 Key takeaway

👉 Meta showed that markets:

• strongly penalize uncertain capex
• but reward execution + capital discipline

📦 3. AMAZON (two comparable cycles)

🔁 AMAZON 2014–2016 (AWS investment phase)

📊 Context:

• Heavy reinvestment in:
  • AWS
  • logistics

👉 Accounting profits were minimal

📉 Market reaction:

• High skepticism:
  • “Amazon will never generate profits”

🔄 Outcome:

• AWS became a:
  • high-margin growth engine

👉 Structural re-rating followed

🔁 AMAZON 2022 (post-pandemic adjustment)

📊 What happened:

• Overinvestment in:
  • logistics capacity

👉 Reduced efficiency and profitability

📉 Market reaction:

• Stock fell ~50%

🔄 Resolution:

• Cost cuts
• Operational optimization

👉 Recovery followed

🧬 4. DIRECT COMPARISON WITH TENCENT

🧠 5. CRITICAL DIFFERENCE (VERY IMPORTANT)

This is the key variable for your investment thesis:

✅ Meta / Amazon

• Operate in:
  • relatively free markets
  • low political/regulatory risk

⚠️ Tencent

Has additional structural risks:

1. China regulatory environment

• gaming
• data
• fintech

2. Intense domestic competition

• ByteDance
• Alibaba
• Baidu

3. Structural valuation discount

• China equity risk premium

👉 Result: The market tends to: apply a deeper and more persistent discount

📊 6. Where inefficiencies may arise

This fits perfectly with your original framework:

👉 Typical opportunity setup:

✅ Signals to look for:

• FCF remains strong (even if temporarily reduced)
• Core business remains stable (gaming + ads)
• Price decline exceeds actual fundamental deterioration

🔴 Warning signs:

• uncontrolled capex expansion
• structural decline in ROIC
• erosion of competitive moat

🧪 7. Quantitative interpretation

Historically in these cases:

👉 The market tends to:

• overestimate the duration of the negative impact
• underestimate the optionality of reinvestment

👉 Result:

• temporary undervaluation (typically 6–24 months)

🎯 8. CONCLUSION (direct and actionable)

Tencent today is in the phase:

👉 Meta 2022 phase (NOT yet in the Meta 2023 recovery phase)

📌 What the market believes:

• AI investments will destroy near-term value

📌 What could actually happen:

• strong execution → major re-rating
• weak execution → long-term value trap

🚀 Final insight for your strategy

This type of situation is historically where significant alpha is generated:

👉 When:

• a “cash machine” company
• shifts into aggressive reinvestment mode
• and the market loses visibility

🧾 2. Can we infer the economic return on Tencent's Ai? NO why?

✅ A. Scale of AI investment

Tencent is entering a phase of combined capex + opex expansion:

• ~RMB 18B invested in AI in 2025
• → targeting ~RMB 36B annualized in 2026

👉 This already includes:

• large language models (LLMs)
• infrastructure (GPUs, ASICs)
• AI talent hiring

✅ B. Quantified impact on profitability (very important)

One of the most relevant — and under-discussed — disclosures:

👉 Tencent has started to separate AI impact in its P&L

• Profit excluding AI:
  • +17% YoY
• Profit including AI:
  • +9% YoY

👉 Difference:

➡️ ~RMB 8.8B quarterly drag from AI ➡️ ~RMB 36B annualized

✅ C. What that number actually represents

This is NOT just capex.

👉 It is largely operating cost (opex):

• model training
• inference (critical)
• AI talent compensation
• energy / data center costs

💡 Key point: 👉 We are already seeing the recurring cost structure of AI

✅ D. Tencent’s implicit economic model

Tencent is effectively doing:

1. Using its core business (gaming + ads)
2. Generating strong FCF
3. Internally funding AI expansion

👉 Implication:

• No immediate financial stress
• But clear pressure on capital efficiency

🧠 2. What we do NOT know (critical gaps)

This is where the real uncertainty lies:

❌ A. No clear disclosure of inference costs

👉 This is THE most important number in modern AI economics:

• Cost per AI interaction
• Cost per agent deployment (e.g., WorkBuddy)

👉 Without this:

• you cannot model scalability accurately

❌ B. No unit economics per AI product

We lack visibility on:

• CAC vs LTV
• incremental margins per AI user

❌ C. No quantified labor substitution

Tencent claims:

• productivity gains
• cost reductions

But does NOT disclose:

• headcount reduction
• salary savings

👉 This is critical for my thesis

❌ D. Limited transparency on cloud + AI margins

We know:

• cloud reached profitability

But we do NOT know:

• margins on:
  • training workloads
  • inference workloads
• internal GPU cost structure

⚠️ 3. Critical analysis (the key section)

🔴 1. The core risk: inference economics

In AI systems:

👉 training = upfront cost 👉 inference = recurring cost

💡 Problem:

• Unlike traditional software:
  • every interaction has a marginal cost

In Tencent’s context:

• ~1.4B users (WeChat ecosystem)
• AI being embedded into:
  • ads
  • messaging
  • productivity tools

👉 If each interaction carries cost:

➡️ The system could become structurally expensive

🧨 2. Does AI actually replace human labor?

This is the core question.

✅ Yes, in some areas:

• customer support
• basic moderation
• content generation
• partial coding tasks

❗ But Tencent’s strategy is NOT primarily cost-reduction

👉 It is more about:

• increasing engagement
• increasing revenue

Example:

• AI-driven advertising improving growth

🔴 Key economic issue:

👉 The cost savings from labor may be smaller than:

• compute costs
• infrastructure costs
• AI talent costs

📉 3. Empirical evidence (very important)

The most revealing data point:

👉 AI reduces profit growth from 17% → 9%

This implies:

👉 As of today, AI is:

• NOT efficiency accretive
• but margin dilutive

🧠 4. Structural shift vs traditional software

Historically:

• marginal cost ≈ zero
• extreme scalability

Now (AI-driven):

• marginal cost > 0
• depends on:
  • compute
  • energy
  • hardware

👉 This fundamentally changes the economics

⚖️ 5. Is human → AI substitution economically viable?

👉 Honest answer: “case-dependent”

✅ Viable when:

• repetitive tasks
• high volume
• low inference cost
• high labor cost savings

Examples:

• customer service
• basic moderation

🔴 Not viable (yet) when:

• complex reasoning
• high precision requirements
• heavy compute usage

Examples:

• advanced AI agents
• deep reasoning systems

📊 6. Final interpretation (investment insight)

👉 Tencent is currently in a:

“AI = capex + opex expansion phase”

NOT in:

“AI = cost reduction phase”

🔑 Critical insight

This is NOT:

“AI reduces costs → margins expand”

This IS:

“AI increases costs → hoping revenue grows enough to compensate”

🔥 This completely changes the DCF profile:

• FCF ↓ (short term)
• capex ↑
• margins ↓
• optionality ↑

🎯 CONCLUSION (direct)

✅ What we know:

• AI cost ≈ RMB 36B annually
• direct negative impact on profits
• real pressure on margins

❌ What we don’t know:

• true unit economics
• cost per inference
• real labor cost savings

🧠 Critical judgment

👉 As of now:

• Tencent’s AI strategy is NOT clearly economically profitable at the operating level
• It is in a phase of:
  • strategic investment
  • long-term optionality capture

🚀 Translation into your framework

If you’re looking for:

“companies where AI replaces labor and improves margins”

👉 Tencent is NOT there yet

👉 It is closer to: Meta 2022 than Microsoft 2015