Tech giants Amazon, Google, Microsoft and Meta are ramping up their investments in artificial intelligence, with a total of more than $320 billion expected to be spent this year.

Despite surprising competition from China's DeepSeek, these companies are convinced that AI will transform many applications and increase demand for cloud services. Executives such as Amazon CEO Andy Jassy emphasize that AI represents a massive opportunity for innovation and growth.

AI is the most significant opportunity since cloud computing and probably the biggest technological change and business opportunity since the internet," Jassy said.

To the surprise of many Wall Street analysts, the big tech companies are sticking to their optimistic growth forecasts for the AI industry.

DeepSeek caused Nvidia's share price to plummet 17 percent, or about $600 billion, last week. The Chinese company developed an AI model similar to Google and OpenAI, apparently without using Nvidia's advanced graphics processors.

Why this matters:.

• Google plans to invest around 75 billion Dollars this year. (53 billion in 2024).
• Microsoft CEO Satya Nadella says his company will invest $80 billion in AI this year.
• Amazon plans to invest about $100 billion in AI. In 2024 it was $77 billion.
• Facebook invested about $40 billion in AI last year. According to CEO Mark Zuckerberg, his company plans to spend “hundreds of billions” on AI. Zuckerberg did not give a timeframe.
• OpenAI's Sam Altman has partnered with SoftBank and Oracle to invest $100 billion in AI infrastructure in the US, although this could rise to half a trillion over time.
• The Japanese investor is currently in talks to invest $25 billion in the start-up, which is valued at $260 billion.

Read on, my dear:

• Wall Street Journal: Tech Giants Double Down on Their Massive AI Spending
• Financial Times: Big Tech lines up over $300bn in AI spending for 2025

Safe Superintelligence: Sutskever's secretive AI startup targets $20bn valuation

Safe Superintelligence (SSI), an AI startup co-founded last year by former OpenAI chief scientist Ilya Sutskever, is in early talks to raise $20 billion in funding, quadrupling the company's valuation from $5 billion in September. Although the company has yet to generate any revenue, SSI's goal is to develop a "safe superintelligence" that exceeds human capabilities while serving human interests.

SSI's secrecy and ability to attract investors is largely due to Sutskever's stellar reputation. He championed the concept of "scalability", which led to the breakthrough of generative AI models such as ChatGPT and triggered massive investments in computing power. Sutskever also recognised the diminishing returns of ever-larger data sets and shifted the focus to the inference phase of AI.

Co-founded by Daniel Gross (former Apple AI executive) and Daniel Levy (former OpenAI researcher), SSI is committed to "scaling in silence", rejecting the rapid commercialisation path taken by some labs, including OpenAI. Little is known about SSI's research direction, except that Sutskever describes it as "a new mountain to climb".

Why it matters:

• A potential valuation of $20 billion suggests growing investor confidence in AI start-ups and highlights the strong market demand for innovative AI technologies.
• With Ilya Sutskever at the helm, a key figure behind major AI breakthroughs such as ChatGPT, the startup's strategy and research carry significant weight in the AI community.
• SSI's goal of "scaling in peace" contrasts with other labs struggling to monetise quickly, suggesting a different model for AI development and commercialisation.

Read on, my dear:

• reuters: OpenAI co-founder Sutskever's SSI in talks to be valued at $20 billion, sources say
• the decoder: Former OpenAI scientist Sutskever's superintelligence startup could hit $20B valuation
• techcrunch.com: Ilya Sutskever’s startup in talks to fundraise at roughly $20B valuation

AI Tip: The Best AI Voice Agents

• Deepgram : Deepgram's speech AI platform provides APIs for speech-to-text, text-to-speech, and full speech-to-speech voice agents. Over 200,000 developers use Deepgram to create speech AI products and features.
• Whisper (OpenAI): Whisper is an automatic speech recognition (ASR) system trained on 680,000 hours of multilingual and multitasking monitored data from the Internet.
• Descript: Descript is an AI-powered, full-featured, end-to-end audio and video editor.

Photo-Prompt:

A curious fluffy cat sitting next to a pastel-colored Easter basket, playfully pawing at a ribbon. Soft sunlight highlights the cat’s fur and the delicate decorations. Realistic photo, natural light

by @han3624 via midjourney

Meta allegedly trained its AI on millions of pirated books

Recently leaked emails reveal details of a copyright lawsuit accusing Meta of training its AI on millions of pirated books. According to the authors' lawsuit, Meta has downloaded at least 81.7 terabytes of data from shadow libraries such as LibGen and Z-Library. Some of these files are said to have been torrents, indicating possible copyright infringement.

The plaintiffs argue that the new revelations contradict previous statements and would require new depositions, including questioning CEO Mark Zuckerberg about his alleged involvement.

Meta denies sharing pirated content with third parties.

Read on, my dear:

ars technica: Torrenting from a corporate laptop doesn’t feel right”: Meta emails unsealed

AI News - What else matters:

Google AI reportedly outperforms gold medallists at International Mathematics Olympiad.
DeepMind's new AI, AlphaGeometry2, has outperformed the average gold medalist in solving geometry problems at the International Mathematical Olympiad. It solved 42 out of 50 () geometry problems, demonstrating advanced problem-solving skills. This hybrid AI combines neural networks with symbolic reasoning and represents a promising approach for the future development of AI.

Hugging Face clones OpenAI's DeepResearch in 24 hours.
Hugging Face has unveiled "Open Deep Research", an open source AI research agent developed in 24 hours in response to OpenAI's Deep Research. This system uses an "agentic" framework together with existing large language models to independently collect and synthesise information, and achieved a score of 55.15% on the GAIA benchmark (compared to 67.36% for Deep Research).

Future plans include supporting more file formats, image-based web browsing, and eventually cloning OpenAI's "Operator" for more extensive tasks. The code is freely available on GitHub, and new contributors are already joining to extend the project's capabilities.