
Meta’s CEO Mark Zuckerberg vowed to double down on investments in AI — even “hundreds of billions of dollars” over the long term
Meta’s upcoming Llama 4 would have multimodal and agentic capabilities and the company's goal this year is to make it the most advanced and widely used AI model
Indian is planning to build a large language model of its own; the government has selected 10 firms, including Jio Platforms and Tata Communications, to supply GPUs for development of machine learning tools needed to build a foundational model
Amid fears that DeepSeek’s AI models could hurt demand for graphics processing units (GPUs), tech giant Meta’s CEO Mark Zuckerberg has vowed to double down on investments in AI — even “hundreds of billions of dollars” over the long term.
During the company’s fourth-quarter earnings call yesterday (January 29), Zuckerberg said he expects Meta’s large language model Llama to become the most advanced and widely used AI model in 2025.
“Our goal with Llama 3 was to make open source competitive with closed models, and our goal for Llama 4 is to lead,” Zuckerberg said.
He further said Meta’s upcoming AI model Llama 4 would have multimodal and agentic capabilities that would unlock several new use cases.
When asked about the impact of Deepseek on Meta’s AI-fueled spending plans, the billionaire said that investing heavily in AI infrastructure will give the company a strategic advantage over time.
While Meta sees Chinese AI startup DeepSeek as a competitor and is learning from it, its founder believes it’s still too early to determine if demand for chips would take a hit as they are crucial for interference purposes.
Zuckerberg said last week that Meta anticipates its capex to be around $60-65 Bn this year, led by investment in AI data centres, compared to $39.23 Bn in 2024.
The company is building a 2 GW AI data centre to bring 1 GW of capacity online this year, closing out 2025 with 1.3 Mn GPUs.
This comes at a time when India is planning to build an LLM model of its own as part of the INR 10,370 Cr IndiaAI Mission.
Union IT minister Ashwini Vaishnaw announced today (January 30) that the government has selected 10 companies that will supply 18,693 GPUs to be used in the development of machine learning tools needed to build a foundational model.
The companies include Hiranandani Group-backed Yotta, Tata Communications, Jio Platforms, and Orient Technologies among others.
After months of discussions with startups and researchers, the government has called for proposals for building India’s own LLM, which is likely to take about 4-8 months, Vaishnaw said.
However, he did not specify how much it would cost to build the foundational model.
Last week, investors were caught off guard after DeepSeek said that its AI models were built far more cheaply than AI systems used by US tech behemoths like Meta and OpenAI. The pessimism roiled shares of tech majors NVIDIA, Alphabet and Microsoft.
However, there is a silver lining in all of this for Indian startups looking to build a homegrown LLM because of the low development costs of DeepSeek’s AI models. According to Perplexity’s founder and CEO Aravind Srinivas, India can build cheaper AI models through ingenuity and strategic investments in open-source models.
Government initiatives like IndiaAI Mission, AI For Agriculture and Future Skills are expected to position India as a global AI leader and an attractive destination for global investors. As per Inc42’s ‘The Rise of India’s GenAI Brigade’ report, homegrown startups are increasingly integrating AI, particularly generative AI (GenAI) into their products and services.
India’s GenAI market is currently valued at $1.6 Bn and is expected to reach a size of $17 Bn by the end of the decade. This will further drive the demand for semiconductor chips. The government on its part has been trying to woo tech giants to set up electronics and semiconductor chip manufacturing units as well as data centres in the country.
Vaishnaw last week said that the first made-in-India chip will be rolled out this year.