Google Touts Flashy Plan for Cheaper AI as Capital Spending Reaches $45 Billion
In the meantime, its sales rose 24% in the second quarter as search revenue climbed 17% and its AI-boosted cloud biz jumped more than 80%.
Sign up for smart news, insights, and analysis on the biggest financial stories of the day.
Google-parent Alphabet hopes money coming in could cool investors’ concerns about money going out.
Its sales rose 24% in the second quarter as search revenue climbed 17% and its AI-boosted cloud biz jumped more than 80%. The Mountain View, Calif.-based company said in April it plans to ramp capital spending to as much as $190 billion this year. In the most recent quarter, its spending tracked toward that, at an expected $45 billion.
Together, the biggest hyperscalers are expected to spend upward of $750 billion this year, and pressure is building for that money to yield results.
The Future Is Flash-Frozen
Hyperscalers are feeling the heat not only from the most advanced AI but also the most affordable, with Chinese AI companies leading the race to the bottom. Alibaba previewed a new version of its Qwen AI model over the weekend, saying the model’s power is second only to Anthropic’s Fable 5. Beijing’s Moonshot AI, meanwhile, debuted Kimi K3, an AI model the developer said outperforms most US systems. Both Chinese models cost significantly less to run than their US rivals.
Before reporting earnings yesterday, Google made moves to cut its own costs:
- On Tuesday, it unveiled three cheaper, lightweight versions of its Gemini AI models. Gemini 3.6 Flash uses as much as 17% fewer tokens, Google said, while costing less per token. An even more threadbare version of its flagship model is Gemini 3.5 Flash-Lite, the least expensive. The third, Gemini 3.5 Flash Cyber, is geared toward cybersecurity.
- Another development that could help Google compete with cheaper Chinese AI models is a new chip that The Information reported the tech giant is developing. Frozen V2, the chip’s codename, could make Gemini up to 10 times more efficient.
Notable Absence: Sideshows can’t distract from a missing main act forever. The tech giant has delayed the launch of Gemini 3.5 Pro, its most powerful model, Bloomberg reported. The model hasn’t been able to bridge the gap between its capabilities and those of the latest models from Anthropic and OpenAI, particularly when it comes to coding. In the AI race, time is money, and Google risks falling further behind.
