One quick thing: PB Fintech’s net profit jumps 92% to Rs 72 crore in Q3, revenue up 48%

In today’s newsletter:

• India joins global LLM race with 18,693 GPUs
• SEBI delivers a deathblow to finfluencers
• Meta bets billions on AI, shrugs off DeepSeek threat

Was this newsletter forwarded to you? You can sign up for Tech3 here

The AI landscape is shifting dramatically.  

First, the USA announced AI export controls. Then, a seemingly unassuming startup, DeepSeek, from China, with its efficient and low-cost generative AI model, disrupted the tech market. Now, it's India's turn. 

Aap chronology samajhiye 

Under the IndiaAI Mission, India is set to develop multiple foundational generative AI models, within the next six - 10 months, said IT minister Ashwini Vaishnaw.

• The Ministry has already identified around six startups to spearhead this development

The foundational model will be designed to address the country’s linguistic and cultural diversity while mitigating biases in datasets, Vaishnaw said.

According to Vaishnaw, creating a common computing facility is the cornerstone of a strong AI ecosystem. To this end, the government has already secured a substantial amount of GPUs:

• Nvidia H100 - 12,896
  
  
• Nvidia H200 - 1,480
  
  
• AMD Mi325x - 512
  
  
• AMD Mi300X - 230

But this is just the beginning. The government is set to expand its GPU procurement even further, with an official notice expected as early as tomorrow.

In addition to ramping up infrastructure, the government is also focused on attracting Indian-origin researchers back to the country.

“Once we create an environment for innovation, Indian-origin developers may be encouraged to return and contribute to building something significant in India,” a government official told us.

Meanwhile, all eyes are on the recent buzz surrounding DeepSeek, a Chinese startup offering an open-source generative AI model that has disrupted the market. 

• To address privacy concerns and safeguard Indian data, the Indian government has assured that the DeepSeek model will be hosted on Indian servers, ensuring that sensitive data remains within the country

The party's over for finfluencers pushing hot stock tips disguised as "education."

SEBI, the market watchdog, has banned financial influencers from using real-time or recent stock prices in their content.

• “Stock market educators” must now use three-month-old data, preventing them from giving real-time trading tips 

This delivers a major blow to those who operated as unregistered investment advisors under the guise of education.

• Without live price discussions or real-time analysis, many may struggle to engage their audience like before 

Finfluencers have blurred the line between education and advice, making it hard to tell who’s credible and who’s just selling dreams.

Also read: How SEBI decisively ends brokerages' romance with finfluencers

Sebi is also tightening its grip on collaborations between registered firms and unregistered finfluencers. 

• No more sponsorships, ads, or paid partnerships
  
  
• Investor education is still allowed, but educators cannot–offer investment advice without registration and make performance claims unless Sebi permits it.

Sebi first restricted such associations in an October 2024 circular, and this latest update on January 29 tightens the rules even further.

With no more loopholes to exploit, many may soon find their influence fading.

Dig deeper

DeepSeek's rapid rise has sent Silicon Valley and Wall Street into panic mode - except for Meta chief Mark Zuckerberg, who remains unfazed.

During Meta's earnings call today, Zuckerberg acknowledged DeepSeek's innovations in building frontier AI models but positioned it as part of the rapidly growing technology's evolution.

"Every new company that has a launch is going to have some new advances that the rest of the field learns from" he said, adding that Meta will implement some of these advances into its own systems.

DeepSeek gained prominence for its claims of building a model that can rival top-tier models from tech giants such as OpenAI, Meta and Google at a fraction of the cost. 

• This sparked concerns among investors about the billions tech giants are investing in developing their cutting-edge AI models and products

Zuckerberg tried dispelling these concerns, saying that investing very heavily in CapEx and infra is going to be a strategic advantage over time. 

• The social networking giant plans to invest "hundreds of billions of dollars" in AI infrastructure over the long term, he said

Also read: Meta to spend up to $65 billion for big AI push in 2025

• 2025 will be the year when a highly intelligent and personalised AI assistant reaches more than 1 billion people, and Zuckerberg hopes Meta AI to be that assistant
  
  
• A pivotal year for Meta since the trajectory for most of the company's long-term initiatives will be clearer by the end of the year
   
• Threads now has 320 million monthly active users, up from 300 million in December
   
• No noticeable impact from the company's recent content policy changes on advertiser spend
  
  
• 2025 will be a big year for redefining Meta's relationship with governments. (The firm is paying $25 million to settle lawsuit over Trump’s account suspension)

Anthropic CEO Dario Amodei has also doubled down on the need for stronger US export controls over AI chips to Beijing against the backdrop of the Chinese AI lab DeepSeek's recent breakthrough. 

Amodei said that DeepSeek's recent model is not a unique breakthrough or a shift in LLM economics but an expected point on a cost-reduction curve. 

• He argued that DeepSeek's total spend is not vastly different from AI labs in the US

The LLM arena just got hotter!

Alibaba just dropped Qwen 2.5, their latest large language model, and they're claiming it outperforms GPT-4, Llama 3, and DeepSeek-V3.

• It boasts multilingual support, enhanced multimodal capabilities (think video and image understanding), and some serious processing power

Learn more in our explainer

• SoftBank is in talks to invest $25 billion directly in OpenAI, a move that could make the Japanese conglomerate the AI startup’s largest financial backer.
  
  
• Khosla Ventures has invested $8 million in C the Signs, a UK startup using AI to predict cancer risk, as the company aims to expand to the US and seeks FDA approval.

Ever wonder how digital devices shape who we are? 

• Analog(ue) digs into how our gadgets affect not just our routines, but our emotions too. 

It’s a deep dive into how tech fits into our daily lives—and how we can find balance in an always-connected world. 

Check it out

Note: By subscribing to Tech3 you have already made the right choice. Top it up with a premium offering, the Moneycontrol Pro Panorama, newsletter that gives you a sharp take on macros, markets, business and finance. Sign up for Pro from this link to get this newsletter in your inbox and also a host of content enjoyed by over a million subscribers.