🤖 Inference cost is the gate on AI search — 10x too expensive today
Friedberg's back-of-envelope: a ChatGPT-style answer costs ~30 cents of compute against ~2.5 cents of cost on a traditional Google search that grosses ~5 cents of revenue — an order of magnitude gap that would run ~$80B a quarter at Google's query volume versus the $8-20B Google actually spends. Chat-native search therefore cannot economically displace keyword search until inference cost falls ~10x, which happens by riding the parallelized-silicon and cheap-energy curves over the next few years. The money in the AI search war accrues to the accelerated-compute and cloud supply chain that closes that gap, not to the search interface.
0
CONVICTION
peaked 68.4 GREEN
EMBER
band
▲ LONG NVDA
expression · bullish tech
HIT
outcome · R +509.9% · α +459.2%
2025-02-11
window closed
⚠ CONFLICTING IDEAS ON THE SAME TICKER
— the board is arguing with itself; net it before trading (ticker view →)
also touching these tickers, same direction: AI compute capex supercycle has 12-24 months of runway 87 (NVDA) · AI compute capex supercycle has 12-24 months of runway 87 (SMH) · SaaSpocalypse is overdone for compliance-moat enterprise software 71 (MSFT) · Nvidia + Hugging Face makes NVDA the open-source AI champion 58 (NVDA) · Nvidia + Hugging Face makes NVDA the open-source AI champion 58 (SMH) · DRAM/HBM: the bottleneck that matters 45 (SMH) · Frontier AI labs keep durable premium pricing (not commoditized) 27 (MSFT) · Own AWS, Azure and GCP and nothing else — the clouds capture the AI application dollars 21 (MSFT) · Independent frontier-model labs lose to hyperscaler capital 14 (MSFT)
⚖ Why this verdict
fully deterministic — evaluate.py replays this from daily closes; nothing below is editable or hand-set
- Window: 2023-02-11 → 2025-02-11 — first mention + 24-month horizon, then the window locks.
- The call: ▲ LONG NVDA (primary play). NVDA rose 509.9% over the window → direction-adjusted R = +509.9% (the call made money).
- Benchmark: holding SPY over the same window returned +50.8% → α = +509.9 − (+50.8) = +459.2% — what this call made or lost against just owning the index. This is the number the verdict uses.
- Rule fired:
▶ HIT — R ≥ +10% AND α ≥ +5 ✓ (R 509.9, α 459.2)· PARTIAL — R ≥ +5% OR α ≥ 0· MISS — everything else
- Credit: supporters of a HIT earn 1.0 each, opposers the inverse — this feeds the scoreboard weights. supported: Chamath, Friedberg, Jason | proxy-sensitive: MSFT→PARTIAL (+54.1%)
Conviction timeline
bands: green ≥ 65 · watch ≥ 45 · ember ≥ 15
Plays vs SPY · % since first mention (2023-02-11)
Plays
| expression | symbol | kind | relevance | rationale |
|---|---|---|---|---|
| ▲ LONG | NVDA | stock | PRIMARY | the specialized parallel silicon that has to deliver the 10x inference-cost reduction Chamath and Friedberg both say is required |
| ▲ LONG | MSFT | stock | adjacent | Azure is where the $10B OpenAI compute commitment gets spent and optimized |
| ▲ LONG | SMH | etf | adjacent | broad semis exposure to the inference cost curve if the winner is not NVDA alone |
Mention log
Friedberg
tech w=1.08 · n=65
· E115 (2023-02-11) · explicit_prediction · strength 3
▶ 38:59
SUPPORT
“A recent estimate on running the GPT-3 model for ChatGPT is that each result takes about 30 cents of compute. So it's about an order of magnitude higher cost to run that search result than it is to do it through a traditional search query.”
Chamath
tech w=0.90 · n=83
· E152 (2023-11-03) · explicit_prediction · strength 2
▶ 1:24:27
SUPPORT
“There's a level of investment multiplied by a level of improvement in the underlying hardware, multiplied by a level of improvement in the underlying models that's creating a 400x improvement from the baseline every year.”
SA
Sam Altman
(guest ×0.5)
no tech track record
· E178 (2024-05-10) · explicit_prediction · strength 2
▶ 6:15
SUPPORT
“We want to like cut the latency super dramatically. We want to cut the cost really, really dramatically.”
Who built this conviction
each voice's total force on the score — supports and opposes from every mention, weighted exactly as the replay applied them · share = % of all mention-driven movement
⏳ decay drained -81.4 over the idea's life — that's time passing, attributed to no one
Score events
| episode | kind | Δ | after | note |
|---|---|---|---|---|
| E115 2023-02-11 | init | +45.2 | 45.2 | E115 born by Friedberg (explicit_prediction x3) [w=1.08] |
| E115 2023-02-11 | reinforce | +13.0 | 58.2 | E115 Jason support x2 (new voice) [w=1.05] |
| E115 2023-02-11 | reinforce | +10.2 | 68.4 | E115 Chamath support x3 (new voice) [w=0.90] |
| E116 2023-02-17 | decay | -1.4 | 67.0 | E116 silent |
| E118 2023-03-03 | decay | -1.3 | 65.7 | E118 silent |
| E119 2023-03-11 | decay | -1.3 | 64.4 | E119 silent |
| E120 2023-03-17 | decay | -1.3 | 63.1 | E120 silent |
| E121 2023-03-24 | decay | -1.3 | 61.8 | E121 silent |
| E122 2023-03-31 | decay | -1.2 | 60.6 | E122 silent |
| E123 2023-04-07 | decay | -1.2 | 59.4 | E123 silent |
| E124 2023-04-14 | decay | -1.2 | 58.2 | E124 silent |
| E125 2023-04-21 | decay | -1.2 | 57.0 | E125 silent |
| E126 2023-04-28 | decay | -1.1 | 55.9 | E126 silent |
| E128 2023-05-12 | decay | -1.1 | 54.8 | E128 silent |
| E129 2023-05-19 | decay | -1.1 | 53.7 | E129 silent |
| E130 2023-05-26 | decay | -1.1 | 52.6 | E130 silent |
| E131 2023-06-02 | decay | -1.1 | 51.6 | E131 silent |
| E132 2023-06-10 | decay | -1.0 | 50.5 | E132 silent |
| E133 2023-06-16 | reinforce | +8.0 | 58.5 | E133 Friedberg support x2 [w=1.08] |
| E134 2023-06-24 | decay | -1.2 | 57.3 | E134 silent |
| E135 2023-07-01 | decay | -1.1 | 56.2 | E135 silent |
| E136 2023-07-09 | decay | -1.1 | 55.1 | E136 silent |
| E137 2023-07-14 | decay | -1.1 | 54.0 | E137 silent |
| E139 2023-07-27 | decay | -1.1 | 52.9 | E139 silent |
| E140 2023-08-04 | decay | -1.1 | 51.8 | E140 silent |
| E141 2023-08-11 | decay | -1.0 | 50.8 | E141 silent |
| E142 2023-08-18 | decay | -1.0 | 49.8 | E142 silent |
| E143 2023-08-25 | decay | -1.0 | 48.8 | E143 silent |
| E144 2023-09-01 | decay | -1.0 | 47.8 | E144 silent |
| E146 2023-09-22 | decay | -1.0 | 46.9 | E146 silent |
| E147 2023-09-29 | decay | -0.9 | 45.9 | E147 silent |
| E148 2023-10-07 | decay | -0.9 | 45.0 | E148 silent |
| E149 2023-10-13 | decay | -0.9 | 44.1 | E149 silent |
| E150 2023-10-20 | decay | -0.9 | 43.2 | E150 silent |
| E151 2023-10-27 | decay | -0.9 | 42.4 | E151 silent |
| E152 2023-11-03 | reinforce | +7.8 | 50.2 | E152 Chamath support x2 [w=0.90] |
| E156 2023-12-08 | decay | -1.0 | 49.2 | E156 silent |
| E157 2023-12-16 | decay | -1.0 | 48.2 | E157 silent |
| E158 2023-12-23 | decay | -1.0 | 47.2 | E158 silent |
| E159 2023-12-29 | decay | -0.9 | 46.3 | E159 silent |
| E160 2024-01-06 | decay | -0.9 | 45.3 | E160 silent |
| E161 2024-01-13 | decay | -0.9 | 44.4 | E161 silent |
| E162 2024-01-19 | decay | -0.9 | 43.6 | E162 silent |
| E163 2024-01-26 | decay | -0.9 | 42.7 | E163 silent |
| E164 2024-02-02 | decay | -0.9 | 41.8 | E164 silent |
| E165 2024-02-09 | decay | -0.8 | 41.0 | E165 silent |
| E166 2024-02-16 | decay | -0.8 | 40.2 | E166 silent |
| E167 2024-02-23 | reinforce | +9.7 | 49.9 | E167 Chamath support x3 [w=0.90] |
| E168 2024-03-01 | decay | -1.0 | 48.9 | E168 silent |
| E169 2024-03-08 | decay | -1.0 | 47.9 | E169 silent |
| E170 2024-03-15 | decay | -1.0 | 47.0 | E170 silent |
| E171 2024-03-22 | decay | -0.9 | 46.0 | E171 silent |
| E172 2024-03-29 | decay | -0.9 | 45.1 | E172 silent |
| E173 2024-04-05 | decay | -0.9 | 44.2 | E173 silent |
| E174 2024-04-12 | decay | -0.9 | 43.3 | E174 silent |
| E175 2024-04-19 | decay | -0.9 | 42.5 | E175 silent |
| E176 2024-04-26 | decay | -0.8 | 41.6 | E176 silent |
| E177 2024-05-03 | decay | -0.8 | 40.8 | E177 silent |
| E178 2024-05-10 | reinforce | +8.0 | 48.8 | E178 Chamath support x2 [w=0.90] |
| E178 2024-05-10 | reinforce | +5.8 | 54.6 | E178 Sam Altman support x2 (new voice) [w=0.50] |
| E179 2024-05-17 | decay | -1.1 | 53.5 | E179 silent |
| E180 2024-05-24 | decay | -1.1 | 52.4 | E180 silent |
| E181 2024-05-31 | decay | -1.0 | 51.4 | E181 silent |
| E182 2024-06-07 | decay | -1.0 | 50.3 | E182 silent |
| E183 2024-06-14 | decay | -1.0 | 49.3 | E183 silent |
| E184 2024-06-20 | decay | -1.0 | 48.3 | E184 silent |
| E185 2024-06-29 | decay | -1.0 | 47.4 | E185 silent |
| E186 2024-07-04 | decay | -0.9 | 46.4 | E186 silent |
| E187 2024-07-12 | decay | -0.9 | 45.5 | E187 silent |
| E188 2024-07-19 | decay | -0.9 | 44.6 | E188 silent |
| E189 2024-07-26 | decay | -0.9 | 43.7 | E189 silent |
| E190 2024-08-02 | decay | -0.9 | 42.8 | E190 silent |
| E191 2024-08-09 | decay | -0.9 | 42.0 | E191 silent |
| E192 2024-08-16 | decay | -0.8 | 41.1 | E192 silent |
| E193 2024-08-23 | decay | -0.8 | 40.3 | E193 silent |
| E194 2024-08-30 | decay | -0.8 | 39.5 | E194 silent |
| E195 2024-09-06 | decay | -0.8 | 38.7 | E195 silent |
| E196 2024-09-20 | decay | -0.8 | 37.9 | E196 silent |
| E197 2024-09-27 | decay | -0.8 | 37.2 | E197 silent |
| E198 2024-10-03 | decay | -0.7 | 36.4 | E198 silent |
| E199 2024-10-11 | decay | -0.7 | 35.7 | E199 silent |
| E200 2024-10-18 | decay | -0.7 | 35.0 | E200 silent |
| E201 2024-10-25 | decay | -0.7 | 34.3 | E201 silent |
| E202 2024-11-01 | decay | -0.7 | 33.6 | E202 silent |
| E203 2024-11-08 | decay | -0.7 | 32.9 | E203 silent |
| E204 2024-11-16 | decay | -0.7 | 32.3 | E204 silent |
| E205 2024-11-23 | decay | -0.6 | 31.6 | E205 silent |
| E206 2024-12-07 | decay | -0.6 | 31.0 | E206 silent |
| E207 2024-12-13 | decay | -0.6 | 30.4 | E207 silent |
| E208 2024-12-20 | decay | -0.6 | 29.8 | E208 silent |
| E209 2025-01-04 | decay | -0.6 | 29.2 | E209 silent |
| E210 2025-01-11 | decay | -0.6 | 28.6 | E210 silent |
| E211 2025-01-18 | decay | -0.6 | 28.0 | E211 silent |
| E212 2025-01-25 | decay | -0.6 | 27.5 | E212 silent |
| E213 2025-01-31 | decay | -0.5 | 26.9 | E213 silent |
| E214 2025-02-07 | decay | -0.5 | 26.4 | E214 silent |