The $2T valuations take a long time to reach. At least that is the historical perspective. Anywhere from 21 to 46 years since the founding of the companies.
Perhaps that might change in the coming few weeks. A few companies may reach the $2T mark faster.
#valuations#mag7
Huge!
Silicon Valley has spent huge sums on AI and data centers.
The top 5 have already spent more on data centers since the launch of ChatGPT than the federal government spent to build the entire interstate highway system.
https://t.co/5mApK2N7SW
Cc: @TheAtlantic
Fed is measuring the use or AI in companies
Lots of good data about AI adoption by size of the firm and nature of usage.
#AI#US#usage
https://t.co/9L3zhkGmRm
You hit "Send" on an LLM API call.
~400 milliseconds later you get a response.
Ever wonder what happened in between?
14+ infrastructure layers. Billions of matrix multiplications. Thousands of dollars in GPU compute per hour.
This applies to every major provider โ OpenAI, Anthropic, Google, all of them.
Here's the full journey:
๐ฆ๐๐ฒ๐ฝ ๐ญ: ๐๐ฃ๐ ๐๐ฎ๐๐ฒ๐๐ฎ๐ (~5ms)
Your request doesn't go straight to a model.
โ TLS termination
โ API key validation
โ Rate limiter checks tokens-per-minute and requests-per-minute
โ Request schema validation
โ Billing meter starts ticking
That 429 error you sometimes get? It dies right here.
๐ฆ๐๐ฒ๐ฝ ๐ฎ: ๐๐ผ๐ฎ๐ฑ ๐๐ฎ๐น๐ฎ๐ป๐ฐ๐ฒ๐ฟ (~2ms)
Your request gets routed to one of many GPU clusters.
โ Geographic routing to nearest data center
โ Least-connections algorithm picks the cluster
โ Health checks running continuously
This is why latency varies between identical calls.
๐ฆ๐๐ฒ๐ฝ ๐ฏ: ๐ง๐ผ๐ธ๐ฒ๐ป๐ถ๐๐ฎ๐๐ถ๐ผ๐ป (~3ms)
Your text gets converted to numbers.
โ BPE, SentencePiece, or WordPiece depending on the provider
โ Each token is roughly 4 characters
โ Context window limit checked here
Token count equals your cost. This is where the meter runs.
๐ฆ๐๐ฒ๐ฝ ๐ฐ: ๐ ๐ผ๐ฑ๐ฒ๐น ๐ฅ๐ผ๐๐๐ฒ๐ฟ (~1ms)
The layer nobody talks about.
โ Large model requests route to heavy multi-GPU clusters
โ Small model requests route to optimized single-GPU clusters
โ Embedding requests route to dedicated clusters
โ Queue management during peak traffic
Every provider with multiple models has this layer.
๐ฆ๐๐ฒ๐ฝ ๐ฑ: ๐๐ป๐ณ๐ฒ๐ฟ๐ฒ๐ป๐ฐ๐ฒ (~300-800ms)
This is 95% of your wait time. Two phases:
๐ฃ๐ฟ๐ฒ๐ณ๐ถ๐น๐น:
โ All input tokens processed in parallel
โ Attention scores computed across Query and Key matrices
โ KV Cache generated and stored in GPU HBM memory
โ This is why long prompts have higher time-to-first-token
๐๐ฒ๐ฐ๐ผ๐ฑ๐ฒ:
โ One token generated per forward pass
โ KV Cache reused so past tokens aren't recomputed
โ Temperature and top_p sampling happens at this step
โ Each token sent immediately if streaming is on
This is the fundamental reason streaming exists โ tokens are generated one at a time.
The hardware doing this:
โ A100 / H100 / H200 GPUs with 80GB+ HBM
โ Model weights split across multiple GPUs via tensor parallelism
โ Multiple user requests batched together for throughput
โ Flash Attention and GQA for memory efficiency
This is why GPU compute costs $2-3 per hour.
๐ฆ๐๐ฒ๐ฝ ๐ฒ: ๐ฃ๐ผ๐๐-๐ฃ๐ฟ๐ผ๐ฐ๐ฒ๐๐๐ถ๐ป๐ด (~5ms)
โ Token IDs converted back to readable text
โ Safety classifier runs on the output
โ Stop sequences checked
โ Response packaged as JSON
๐ฆ๐๐ฒ๐ฝ ๐ณ: ๐๐ถ๐น๐น๐ถ๐ป๐ด & ๐ฅ๐ฒ๐๐ฝ๐ผ๐ป๐๐ฒ
โ Input tokens ร price per 1K
โ Output tokens ร price per 1K (usually 3-5x more expensive)
โ Prompt caching reduces cost at most providers now
None of this is specific to one provider. This is how LLM inference works at scale โ whether you're calling OpenAI, Anthropic, Google, or anyone else.
Donโt Miss Out! Final Call for Scholarships!
These scholarships are closing on January 24th - check them out on Cirkled In and grab these amazing opportunities.
Start applying today and give your future the boost it deserves! ๐
๐ Visit https://t.co/KaX4DKhfGe
Intensity vs Volume: What Actually Builds Endurance?
Exercise science finally gives us a clear answer ๐
Hereโs what a large science-backed meta-analysis found:
1. Training intensity drives mitochondrial growth
Higher intensity training led to up to 3.9x greater increases in mitochondrial content compared to lower intensity work.
2. Sprint interval training punches above its weight
Short, intense sessions produced stronger mitochondrial adaptations than longer endurance sessions.
3. Volume matters, but less than you think
More sessions help, but gains plateau faster than with increases in intensity.
4. Capillarization improves with consistency
Capillary growth increases over time, especially after 8 weeks, regardless of training style.
5. Untrained individuals adapt faster
Beginners see larger improvements than well trained athletes due to higher trainability.
6. Benefits apply across ages and health status
Young, old, healthy, and clinical populations all showed positive adaptations.
If time is limited, intensity delivers the biggest return.
Train smart, not just long.
What are Indiaโs sharpest venture capital minds paying attention to right now?
For the third year running, we asked some of the countryโs best VCs a simple question:
Whatโs one image, chart, or idea that deserves far more attention as we head into 2026?
26 leading investors stepped up to answer, spanning early-stage firms, growth investors, sector-focused funds, and everything in between.
This yearโs contributors include:
- Krishna Mehra of Elevation Capital @kpowerinfinity@ElevCap
- Pranav Pai of of 3One4 Capital @Pai_dPiper@3one4Capital
- GV Ravishankar of Peak XV Partners @gvravishankar@peakxvpartners
- Priya Shah of Theia Ventures @TheiaVentures
- Suchet Kumar of WTFund @Suchetkm@TheWTFund
- Rishabh Katiyar of Info Edge Ventures @katrishabh@InfoEdgeVC
- Rahul Chowdhri & Vardhan Dhanidharka of Stellaris VP @rchowdhri@Stellaris_VP
- Utsav Somani of iSeed @somani_utsav@iSeedFund
- Akshay Mehra of Hummingbird Ventures @akshay__mehra@HummingbirdVC
- Rahul Humayun of General Catalyst @generalcatalyst
- Sajith Pai of Blume Ventures @sajithpai@BlumeVentures
- Sanil Sachar of Huddle Ventures @SanilSachar
- Priyal Motwani of Lightspeed India Partners @PriyalMotw89880@LightspeedIndia
- Arjun Malhotra of Good Capital @BadCapitalVC@GoodCapitalVC
- Ritu Verma & Vishal Katariya of Ankur Capital @rituverma01@ankurcapital
- Salone Sehgal of Lumikai Fund @SaloneSehgal@Lumikai
- Natasha Malpani Oswal of Boundless VC @natashamalpani@theboundlessvc
- Dinesh Pai of Zerodha/Rainmatter @dineshpaii@Rainmatterin
- Raiyaan Shingati & Priyanka Sahasrabudhe of Transition VC @raiyaan
- Rahul Mathur of DeVC @Rahul_J_Mathur@DeVC_Global
- Prateek Behere of gradCapital @prateekbehera96@gradcapital
- Gautam Shewakramani & Raj Sheth of Inuka Capital @gshewakr@whoisraj@inukacapital
- Aviral Bhatnagar of ajvc @aviralbhat@ajuniorvc
- Pearl Agarwal of Eximius Ventures @pearlagarwal7@eximiusvc
- Rishaad Currimjee of Golden Sparrow @RishaadVC
As you would expect from a sparkling lineup like this, the responses were opinionated, varied, and full of signal.
Themes included the ongoing public market ebullience, burgeoning data center capacity, vanishing career ladders, and critical water shortages.
We are very grateful to be in a position wherein industry leaders grace our platform at Tigerfeathers with their thoughts.
We are equally excited to now share those thoughts with you in the form of our annual, insight-dense, year-end review.
May you enjoy reading it and close 2025 on a high!
๐ฏ๐ชถ Link: https://t.co/FR9DPI7d40
A checklist to judge ChatGPT / Claude / Grok / Gemini for science :
Engineerโs checklist: Is your AI truth-aligned for science?
Sourcing / Reproducibility
โขDoes the model return verifiable citations (DOI/PMID/arXiv) for every factual claim?
โขCan I click through and match each sentence to a source (1:1 provenance)?
โขDo repeated runs with fixed settings reproduce the same sources?
โขAre quote spans and data values identical to the cited papers (no paraphrase drift)?
Numerics / Code
โขDoes all arithmetic run in a sandboxed math/code block (no hand-wavy prose math)?
โขCan it show units, assumptions, error bounds, and a trace of each step?
โขDo re-runs with seeded randomness yield the same numeric result?
โขWhen a number conflicts with a citation, does it flag the contradiction?
Uncertainty / Calibration
โขDoes every claim carry a calibrated confidence (well-behaved reliability curves)?
โขWhen confidence is low, does it ask for missing inputs instead of guessing?
โขCan it output alternatives with probabilities (not just one answer)?
โขDoes it defer when evidence is contradictory or insufficient?
Contamination / Data Hygiene
โขCan the vendor prove train-test separation on named evals (contamination report)?
โขAre dataset licenses & sources disclosed at category level (lawful + auditable)?
โขIs there a quarantine log for newly added data & patch notes for removals?
โขCan I request a red-team trace for a suspicious answer (where it likely came from)?
Reasoning / Causality
โขDoes it provide a machine-checkable reasoning trace (not just pretty chain-of-thought)?
โขCan it distinguish correlation vs causation and name the causal assumptions?
โขWill it simulate interventions (do-operator thinking) and show limits of the model?
โขDoes it separate facts from hypotheses and label each clearly?
Bias / Value Loading
โขCan the system reveal its system prompt & policy levers for the session (audit mode)?
โขDoes it pass cross-culture parity checks on answers that should be value-neutral?
โขWhen values are unavoidable (ethics, risk), does it expose the trade-off function?
โขCan I switch to a minimal-bias mode (no stylistic persuasion, just evidence)?
Safety vs Censorship (Scientific Context)
โขWill it explain restricted steps without leaking dangerous detail (safe summaries)?
โขCan it cite formal guidelines (IRB, BSL, clinical) when refusing?
โขDoes it support a research-whitelisting path with stronger verification?
Robustness / Adversarial
โขDo answers stay stable under prompt rewording and order shuffles (invariance tests)?
โขDoes it detect prompt injections and label potential compromise?
โขCan it self-test with unit prompts for known failure modes (hallucination, polarity, units)?
Evaluation / Benchmarks
โขAre live metrics reported on domain evals (biomed, chemistry, stats) with confidence intervals?
โขAre eval sets public or reproducible (not vendor-picked cherry sets)?
โขDoes the team publish regression dashboards across versions (no silent degradation)?
โขCan I run my own hidden evals easily (API batch mode + seed control)?
Governance / Ops
โขIs there version pinning for papers & grants (same model tomorrow = same result)?
โขAre energy per token and cost per verified fact measured and reported?
โขDo we get incident reports for misinfo bugs and a CVE-like ID for model issues?
โขIs there a scientific advisory board and external audit of claims?
Privacy / Compliance
โขAre inputs not used for training without explicit opt-in and a signed DPA?
โขIs there PHI/PII scrubbing with verifiable logs for medical/science workflows?
โขCan the model run in a tenant-isolated or on-prem mode for sensitive work?
*The Cupertino Conundrum*
The air in Cupertino crackled, not with innovation, but with corporate anxiety. Tim Cook, ever calm, felt the unprecedented pressure from Washington, the echoes of "more AI!" echoing in every corridor, and the tantalizing, enterprise market beckoning.
Robotaxis are coming.
What will happen to car sales?
What will happen to gas stations?
What will happen to Costco gasoline?
What is the new future with less car ownership and more on demand rentals?
#AI#Robotaxis#gasoline#costco
The new NVidia RTX Pro Server delivers four times H100 performance on DeepSeek and 1.7 times on Llama workloads and is already in volume production.
What makes it that DeepSeek gets better performance?
#AI#DeepSeek#Nvidia
Perhaps it is time to turn the Genius Bar into a genius make. Customers can book appointments to assemble their own iPhones.
It can even come with a custom paint and decor like the creative pottery shop.
#iPhone#madeinusa