thomas

T2 · manager / operator

Thomas is a community member known for deep technical knowledge of AI chip architecture and the semiconductor value chain. He runs a concentrated portfolio (nearly 100% AMD) with small positions in Tesla and ImmunityBio. He uses AI coding agents extensively in his workflow and has a background in self-taught AI/math coursework that led to his investment thesis.

12 calls·12 names·75% bull·last heard 2 months ago·Antonio Linares+1
track recordleaderboard →
hit rate
40%
avg alpha
+6.4pp
scored
5

top calls

best measured alpha vs SPY, then highest conviction · one per company
1st+89.7pp vs SPY
$AMDAdvanced Micro Devicesposition

Thomas goes nearly all-in on AMD citing chiplet flexibility and CPU-GPU integration

AMD's chiplet architecture provides superior modularity for inference workloads, enabling circuit-level optimization that monolithic designs cannot match. The CPU-GPU integration advantage becomes critical as agent architectures shift workload to CPUs, and OpenAI's reported $100B chip deal signals recognition of AMD's strategic position.

2nd+5.3pp vs SPY
$IBRXImmunityBio

ImmunityBio's IL-15 pathway could rejuvenate immune system like Nvidia did for GPUs

ImmunityBio's IL-15 pathway therapy allegedly jacks up T and NK cell counts, rejuvenating the immune system — a foundational biological pathway analogous to Nvidia's role in GPU compute. Thomas considers allocating $200K based on intuitive conviction.

3rdhigh conviction
$HIMSHims & Hers Healthposition

Hims applies Costco model to healthcare: biomarkers + deflationary scale + customer acquisition via acute needs

Hims combines exhaustive biomarker testing (top-of-funnel data) with industrial-scale fulfillment of personalized molecules, acquiring customers through acute needs (ED, hair loss) then funneling into preventive biomarker subscriptions. This bottoms-up model disrupts top-down healthcare.

most discussed · click a bar to filter

12 total
$DUOL
···
Duolingo
HIGHthomas·Antonio Linares·7 months ago
Duolingo seen as potential 5-20 trillion dollar company from AI-driven education transformation
Duolingo's language learning platform will become a primary beneficiary of AI-driven education transformation, potentially reaching multi-trillion dollar valuation as the educational shift accelerates toward personalized AI tutoring.
"I mean yeah that's obvious like I think um yeah I people are very triggered by that. I think for us right now it's a very normal thing to say but I think it's a trillion dollar 5…"
43:21
$AMD
···
Advanced Micro Devices
HIGHthomas·Antonio Linares·7 months ago· position
Thomas goes nearly all-in on AMD citing chiplet flexibility and CPU-GPU integration
AMD's chiplet architecture provides superior modularity for inference workloads, enabling circuit-level optimization that monolithic designs cannot match. The CPU-GPU integration advantage becomes critical as agent architectures shift workload to CPUs, and OpenAI's reported $100B chip deal signals recognition of AMD's strategic position.
"My whole thesis is basically the modularity of the chiplet infrastructure will give AMD at least half the market, right? That's something on the inference market. Yes, inference m…"
4:11
$MU
···
Micron Technology
MEDthomas·Antonio Linares·7 months ago
Micron benefits from HBM shortage as key supplier to AMD and Nvidia
Micron is a primary HBM supplier benefiting from severe shortage driven by AI chip demand; HBM is a commodity with few providers (Micron, SK Hynix, Samsung) creating pricing power.
"yeah Micron uh yeah ... Yeah uh SKHix or Samsung or um some smaller companies but they are those stocks are pumping right now because of the shortage."
10:41
$TSLA
···
Tesla
LOWthomas·Antonio Linares·7 months ago· position
Thomas holds small Tesla position alongside concentrated AMD bet
Tesla held as a minor position complementing the core AMD thesis, likely for its AI/robotics (Optimus) and energy exposure.
"I'm almost full full portfolio in AMD right now. I have bits in Tesla."
17:15
$AVGO
···
Broadcom
MEDthomas·Antonio Linares·7 months ago
Broadcom's TPUs serve Google's Gemini but lack workload flexibility vs GPUs
Broadcom's TPUs power Google's leading Gemini models but cannot easily reconfigure for architecture shifts (e.g., transformer to diffusion), making them less flexible than GPUs for diverse AI workloads.
"you would have the TPU market which is mainly Broadcom or Google TPU which is also you know the same uh and you would also have the new the new players like Grock or uh Cerebras .…"
6:24
$CBRS
···
Cerebras
MEDthomas·Antonio Linares·7 months ago
Cerebras' wafer-scale SRAM architecture targets ultra-low latency inference niches
Cerebras' unique wafer-scale SRAM architecture delivers exceptional inference speed for specific use cases but faces scaling limitations for large models due to memory constraints.
"you would also have the new the new players like Grock or uh Cerebras which are using a lot of SRAMM which is um not HBM so it's it's a different kind of architecture which makes…"
6:24
$IBRX
···
ImmunityBio
HIGHthomas·Antonio Linares·7 months ago
ImmunityBio's IL-15 pathway could rejuvenate immune system like Nvidia did for GPUs
ImmunityBio's IL-15 pathway therapy allegedly jacks up T and NK cell counts, rejuvenating the immune system — a foundational biological pathway analogous to Nvidia's role in GPU compute. Thomas considers allocating $200K based on intuitive conviction.
"I was ranting this week about immunity bio and the 15 pathway and how that allegedly um jacks up T and NK cell count and everything. And going back to the idea of self-actualizati…"
37:13
$GROQ
Groq
MEDthomas·Antonio Linares·7 months ago
Groq's SRAM-based architecture excels at low-latency inference for customer service agents
Groq's SRAM architecture (not HBM) enables extremely fast inference for latency-sensitive applications like customer service agents, though scaling to larger models remains challenging.
"you would also have the new the new players like Grock or uh Cerebras which are using a lot of SRAMM which is um not HBM so it's it's a different kind of architecture which makes…"
6:24
$HIMS
···
Hims & Hers Health
HIGHthomas·Antonio Linares·7 months ago· position
Hims applies Costco model to healthcare: biomarkers + deflationary scale + customer acquisition via acute needs
Hims combines exhaustive biomarker testing (top-of-funnel data) with industrial-scale fulfillment of personalized molecules, acquiring customers through acute needs (ED, hair loss) then funneling into preventive biomarker subscriptions. This bottoms-up model disrupts top-down healthcare.
"I think it's the Costco algorithm just applied to the various components of the healthcare value chain and the ones that's actually fundamental is exhaustive biomarkers. So if if…"
38:27
$TSM
···
TSMC
MEDthomas·Antonio Linares·7 months ago
TSMC capacity constraints limit CPU/GPU supply, creating shortage dynamics
TSMC's finite advanced packaging capacity creates structural bottleneck for AI chips; the key question is whether TSMC, Intel, or AMD captures the pricing power from resulting shortages.
"I'm thinking they're going to there's going to be a CPU shortage because the TSMC and the Intel FAPS are not in infinitely scaling. Um and but the only question right now is so wh…"
12:04