Ter info.
Citi on META (Buy; PT $753): 'DeepSeek R1 Should Lead to Greater GenAI Product Cadence and Potentially Faster ROI'
"Following the launch of DeepSeek’s open-source R1 model and given performance relatively on par with other major models (like Gemini & ChatGPT 4.0), but reportedly trained at a significantly lower cost (~$6M), it should lead to a faster GenAI product cadence delivering greater (and faster) ROI. Key here is R1’s open-source approach as it enables other models—like Llama, Gemini, and OpenAI—to learn from its efficiency gains.
Interestingly, R1’s efficiency gains come on the heels of Meta announcing its plans to spend $60-65B on CapEx this year and adds to our view that Meta’s GenAI investments can lead to greater engagement as newer experiences—like Search, Agents, and Enterprise—launch across multiple form factors. Meta’s 4Q earnings are this Wednesday AMC, and we’ll be listening for whether model efficiency gains could change the trajectory of product (and CapEx) spend, among other benefits."