This week, Google signed a $920 million monthly contract with SpaceX for AI compute capacity, Meta launched its Business Agent globally on WhatsApp and Instagram, OpenAI introduced persistent memory in ChatGPT with Dreaming V3, Huawei completed training of DeepSeek V4-Pro using domestic chips, and Microsoft unveiled seven proprietary AI models to compete directly with OpenAI.
I’ve filtered out the noise for you
This was a week of structural moves in the AI industry. Not minor updates: Google committed over $10 billion annually to SpaceX infrastructure, Meta opened its business agent to every company in the world, and Microsoft began separating from OpenAI with its own models. Each of these stories redefines who controls enterprise AI over the next several years.
1. OpenAI redefined ChatGPT memory with Dreaming V3
OpenAI launched Dreaming V3 on June 4, a persistent memory system that synthesizes context from multiple conversations in the background. Unlike the previous system, it requires no manual reminders from the user. The model builds a cumulative understanding of the user over time. Rollout began with Plus and Pro users in the United States.
2. Google signed a $920M monthly deal with SpaceX for AI compute
Google agreed to pay SpaceX $920 million per month from October 2026 through mid-2029, accessing approximately 110,000 Nvidia GPUs, CPUs, and memory in SpaceX data centers. The capacity serves as a bridge for the Gemini Enterprise agent platform amid unexpected demand. The deal was announced one week before SpaceX’s anticipated IPO, which aims to raise $75 billion at a $1.75 trillion valuation.
3. Meta launched its Business Agent globally on WhatsApp, Instagram, and Messenger
Meta launched its Business Agent globally on June 3, available on WhatsApp Business, Instagram DMs, and Messenger. The agent can answer questions, recommend products, book appointments, qualify leads, and route queries to human agents. It operates across more than one billion daily business-customer conversations. Currently free, with tiered pricing coming soon.
For sales teams tired of cold leads, slow customer responses, and manual processes, Dapta is the ultimate tool.
Dapta is the leading platform for creating AI sales agents specifically designed to increase inbound lead conversion. Respond to your leads in less than a minute with voice AI and WhatsApp that converts.
If you want your team to sell more while AI handles the complex stuff, you have to try it.
4. Huawei completed training of DeepSeek V4-Pro using domestic chips
Huawei completed post-training of DeepSeek V4-Pro, a 1.6 trillion parameter model, using a cluster of at least 1,000 Huawei Ascend 910C chips. This achievement represents a significant step toward Chinese AI independence from US semiconductor export sanctions. Researchers describe the process as considerably more complex than inference, requiring intense coordination across thousands of processing cores.
5. Microsoft unveiled seven MAI models to compete with OpenAI
Microsoft launched seven proprietary AI models under the MAI brand. MAI-Code-1-Flash specializes in programming, while MAI-Thinking-1 was trained from scratch using commercially licensed data, without OpenAI refinement. The launch marks a concrete step toward Microsoft’s technological independence from its most important AI partner.
Bonus: Content of the Week
Single-founder companies are reaching unicorn valuations in Silicon Valley. AI reduces the need for large teams to build scalable products, enabling one person to build and operate what previously required dozens of engineers.
