Mandeep Singh would be surprised if Apple's long-term AI strategy is to rely solely on Google's Gemini model.

Mandeep Singh predicts that Apple will become more acquisitive under its new leadership than it was under Tim Cook.

Mandeep Singh questions why Apple did not increase capital expenditures for AI data centers while its 'Magnificent Seven' peers were making significant investments.

Mandeep Singh suggests that for Apple to acquire a significant AI company like Perplexity or Cursor, it might need to spend between $50 billion and $100 billion.

Mandeep Singh believes Apple will have to acquire AI technology because it has not invested enough in its own internal AI development.

Mandeep Singh believes that relying solely on its distribution advantage for AI is a risky strategy for Apple that could cause it to miss a major secular trend.

Mandeep Singh believes appointing Apple's hardware chief as CEO is a logical decision because the company's $100 billion in free cash flow is primarily driven by its hardware business.

Mandeep Singh believes appointing hardware chief John Ternus as CEO is a logical decision because Apple's business is still fundamentally reliant on hardware.

Mandeep Singh believes Meta's return on investment for expensive AI chips is challenged by its lack of a cloud business to rent out excess capacity, unlike Microsoft, Amazon, or Google.

Anthropic has an Annual Recurring Revenue (ARR) of $30 billion.

Anthropic has surpassed OpenAI and other frontier LLMs in the specific application of coding agents.

AI companies like Anthropic and OpenAI require significant capital infusions because they are compute constrained and lack the free cash flow of hyperscalers to fund CapEx.

OpenAI was able to raise $120 billion before Anthropic surpassed its ARR.

Google is considered the leading company in token efficiency due to its successful implementation of TPUs for both AI training and inference workloads.

According to Mandeep Singh, Google has the highest token efficiency in the industry due to its successful implementation of TPUs for both AI training and inference workloads.

Mandeep Singh believes it is likely too late for Meta to launch a competitive cloud business, an opportunity it should have pursued years ago similar to Oracle.

Meta redesigned its pre-training stack over the last nine months to improve token efficiency.

Anthropic has surpassed OpenAI's Annual Recurring Revenue (ARR).