GPT-6 Astra is framed as a more efficient but more capable model that strengthens the AI buildout trade while shifting the winners toward infrastructure, power, and new application layers. The near-term setup still favors AMZN, OpenAI-linked names, and power plays like VST and BLMN-style beneficiaries, while the longer-term thesis is that AI efficiency will trigger a wave of earnings upside, not less AI spend. Gaming is the clearest software application trade: UNITY is pitched as the picks-and-shovels layer for AI-generated worlds, and publishers with huge universes like T2 gain from lower-cost expansion. The other big call is that the OpenAI/Anthropic cycle remains a duopoly with trillion-dollar IPO/re-rating potential, and that investors are underestimating how much value sits in equity, distribution, and infrastructure around it.
GPT6 Astra launched Friday, scored 99.9% on OpenAI’s Arc AGI benchmark, and is described as the first OpenAI model to hit the company’s critical cyber-security threshold.
Astra is portrayed as more efficient on computer tasks, taking about half the time versus Saul, with the core debate now whether cheaper AI means less spend or 10x cheaper AI driving 100x more usage.
The infrastructure trade is said to stay alive until a major shift away from chips, memory, data centers, and power; Astra is intriguing but "not there yet," so spend still scales with compute.
Open-source models from China are cited as having unraveled the AI trade before, and a new open model is expected in the next few weeks to months to scare investors into thinking OpenAI and Anthropic do not have a shot.
That scare is expected to be followed by an OpenAI or Anthropic model that looks like ASI and reasserts trillion-dollar value, with Anthropic IPO talk centered around $2T to $2.5T and an earmark around $2.5T.
AMZN was added today on a report that Amazon is moving up the AI infrastructure supply chain, increasing order sizes, and accelerating delivery of racks into its data centers.
AMZN is described as having almost perfect visibility into its AI infrastructure activity, with claims of about $100B in OpenAI infrastructure commitments, $15B plus another $35B in OI, and exclusive OpenAI distribution through AWS/Bedrock.
The Anthropic IPO is said to imply AMZN owns 8% to 9% of Anthropic, translating to a couple hundred billion dollars for AMZN, while OpenAI ownership and future IPO upside are framed as another $100B to $150B in cash value.
OpenAI is described as the #1 or #2 most downloaded iOS app globally for 2.5 years, with about a billion people using ChatGPT through the day and consumer adoption called unprecedented.
MSFT, OpenAI, and Anthropic are framed as a duopoly-like frontier-model setup, with weak sentiment masking the scale of the businesses; GOOGL and Gemini are described as looking worse than ever.
SOFTBANK is called the purest way to play the OpenAI/IPO re-rating setup, with the note that SoftBank owns about 13% of OpenAI, has a market cap around $230B, and also owns ARM.
UNITY is a new buy because GPT6 Astra makes one-shot video-game creation cheaper and easier; Unity owns one of the biggest environments where agents can build interactive experiences and has an AI gateway for code inside the editor.
OpenAI’s Astra launch used UNITY as the case study, and with Unity integration via PLCO, OpenAI reported 50% fewer manual fixes versus GPT-5.
UNITY is framed as the picks-and-shovels layer for AI-generated worlds: it can take a cut of gross game revenue, could be an acquisition target for OpenAI, Anthropic, or another frontier model at $30B to $35B, and has an $18B market cap.
Gaming is called ripe for development because budgets are huge and the sector is boom-bust; GTA 6 is singled out as a 10-year universe that could expand with AI while preserving IP and community connectivity.
T2 is said to benefit from AI-enabled universe expansion, while AMZN, OI, VST, MU, and BLMN are all treated as different ways to own the AI efficiency/infrastructure ripple effect.