Reshab Shaw (@shaw_reshab) header
Reshab Shaw (@shaw_reshab) profile picture
Bangalore
Joined February 2019
Photo by Reshab Shaw (@shaw_reshab)Photo by Reshab Shaw (@shaw_reshab)Photo by Reshab Shaw (@shaw_reshab)Photo by Reshab Shaw (@shaw_reshab)Photo by Reshab Shaw (@shaw_reshab)Photo by Reshab Shaw (@shaw_reshab)Photo by Reshab Shaw (@shaw_reshab)Photo by Reshab Shaw (@shaw_reshab)Photo by Reshab Shaw (@shaw_reshab)Photo by Reshab Shaw (@shaw_reshab)
Reshab Shaw retweeted
Some clarifications on various themes being discussed about @SarvamAI in the last few days: 1. Government Support: The IndiaAI Mission under MeitY has played a pivotal role in bootstrapping India's AI ecosystem. It gave ~20 startups subsidised access (40% discount) to GPUs for training models. We used that access to train Sarvam 105B from scratch, which we completed and open-sourced this February. The government did not pay Sarvam directly, instead like with every supported startup, the subsidy went to GPU cloud providers, helping seed AI infra in the country. 2. Sarvam Token Factory: Open-weight models saw massive adoption globally this year. They offer a cost advantage over closed models and let enterprises own their AI stack. But this choice is not available to all Indian enterprises, which often need data to remain within the country and low latency for interactive use cases. Sarvam Token Factory is one of the first large inference offerings in India. We serve multiple open-weight models on GPUs we have privately procured, without any government support. We have optimised datacenter scale inference to make our token cost comparable to or lower than established providers such as Fireworks, Baseten and Together. Today, India consumes a few trillion tokens a day and imports the majority of them. Scaling this consumption up and building sovereign capability requires investing in GPU infra, running efficient token factories, and guaranteeing advantages in cost, latency, and data residency. This is the key objective of Sarvam Token Factory. We will continue to reduce costs further and widen access for inferencing in India. 3. Our own models: Building inference capability does not conflict with our core mission: training our own models. Over the last few months, alongside putting together the token factory and scaling up our enterprise offerings, we optimised our training stack for Blackwell GPUs and achieved significantly higher performance. We used this optimised infra to train an improved checkpoint: Sarvam 105B-1711, which we are open-sourcing today. It improves on reasoning, tool-calling, and co-work tasks, while excelling in Indian languages. It is one of the best Flash-Lite class models, open or closed, achieving high accuracy at a low cost of ₹25 per million output tokens. We also optimised it to run on a single RTX Pro GPU with our agentic product Indus. Along with the weights, we release inference and training recipes so others can build upon it. In the coming two weeks we will release multiple other models across vision, speech, and edge. And crucially, we are working with partners to get the latest training grade compute into India. This will set us up for training our next generation of much larger models. 4. The Sarvam Journey. To put things in perspective, most AI labs invest 100s of millions of dollars into infra to train from scratch, scaling up to billions for frontier capabilities. We are determined to climb that ladder from India, while building a sustainable commercial engine across population scale and enterprise use cases. Irrespective of the different narratives, some of which appear manufactured, we firmly believe that India can and must win across the full-stack of AI - we have the talent, the demand, and positivity to adopt AI. At Sarvam we will stay focused and build towards this outcome.
14
56
3
194
3,413
Reshab Shaw retweeted
everyone’s dunking on sarvam and comparing it to openai and anthropic look at the numbers first: GPUs sarvam: ~4,000 H100s all of india’s AI mission: ~34,000 openai: ~1.7 million anthropic: 1 million TPUs from one google deal alone funding sarvam: ~$275M total anthropic: ~$144B openai: ~$180B that’s roughly 500x more money, and people are still comparing india’s startup culture, AI, healthtech, fintech, it all just arrived, the US and china had a 10-year head start, this isn’t a fair fight yet, it’s chapter one and honestly, the audacity of people sitting comfy in wipro-style jobs, never stepping out of their comfort zone, never building anything for the country, but always first in line to dunk on the ones who are, is so sick my honest take on sarvam though, pick one thing, fine-tune it properly, make it great, stop trying to do everything at once and if the real plan is government and enterprise? go all in, close it for everyone else and own that market
50
28
9
263
19,817
#WATCH | Anthropic CEO Dario Amodei and the company's co-founders ring the opening bell at the Nasdaq exchange on Thursday, marking a major milestone as the maker of Claude AI enters the public markets. Anthropic is listed on the Nasdaq Global Select Market under the ticker $ANT. The listing is among the biggest IPOs in history and marks a significant moment for the AI industry, as investors get their first opportunity to buy shares in one of the leading developers of generative AI models. #Anthropic #DarioAmodei #Nasdaq #IPO #ArtificialIntelligence #WallStreet #Moneycontrol
2
8
1
29
14,625
#TechWithMC | Cognizant beats TCS, HCLTech and two global rivals to win $175-million UK data centre deal -Cognizant wins £133.2m Network Rail tech contract. -Beats rivals TCS, HCLTech, DXC, and Atos for the deal. -Seven-year agreement covers data centre and cloud ops. @shaw_reshab reports: moneycontrol.com/news/busine…
1
3
19
7,788
🚨 Exclusive: Cognizant beats TCS, HCLTech and two global rivals to win $175-million UK data centre deal IT services major Cognizant has emerged as the winner of a £133.2-million (approximately $175 million) technology infrastructure contract from the UK's Network Rail Infrastructure Limited, beating four competitors, including Indian IT majors Tata Consultancy Services (TCS) and HCLTech. The nearly seven-year contract covers the management and operation of Network Rail's data centre infrastructure, including networks, servers, mainframes, cloud platforms, and associated IT services. Break by @shaw_reshab moneycontrol.com/news/busine…
6
20
172
9,690
Really don't understand why launching a token-factory product is getting so much criticism thrown @SarvamAI. I am obviously biased as an investor but I think people don't know just HOW MUCH some of the global players are making by hosting their global (incld chinese) models in India AND serving global customers from here, simply because they can't find GPUs elsewhere right now! You'd be shocked. And yes, goes without saying, the larger models are part of the roadmap, but just because a company launches product-x, doesn't mean that's where the launches are going to stop? There are so many parallel teams working on really ambitious products all launching at their own cadences. Genuinely confusing... All the other feedback around model latency, cost, performance etc are all really useful ones.
106
52
21
699
58,812
Jensen Huang's exit unlikely to derail AI boom as trillions in investments keep Nvidia central - Jensen Huang resigns after over three decades leading Nvidia. - CFO Colette Kress is appointed as interim CEO. - Experts see secure near-term prospects despite the exit. @shaw_reshab & @debanganaghosh4 with more details moneycontrol.com/artificial-…
3
1
12
8,559
Panicked for a second wondering how we missed news of large scale layoffs. Turns out this India Tech handle picked it from IB Times, which in turn curated a story on Cognizant job cuts yesterday from a story we did way back in May 2026, combining it with Qualcomm, which happened this week🧐. Crazy to do this belated curation now and cause widespread alarm. Here's the original link to the story that @shaw_reshab and I broke earlier this year moneycontrol.com/news/busine…
🚨 Cognizant is planning to cut 15,000 jobs globally. (IBTimes)
18
75
2
671
63,563