Ideas
Technology spending will not slow.
The tug of war over AI safety between Anthropic/OpenAI and the rest of the AI food chain will not stop the massive buildout; plenty of money will be spent on technology, and there is no slowdown despite Wall Street worries.
Nvidia is best AI compute investment.
Nvidia is the best investment because AI is now useful and compute demand is exploding; each gigawatt Nvidia AI factory costs $50-60B and generates about $50B in annual rentals with roughly one-year ROI, Nvidia compute is fungible and durable across every model and AI stage, spot GPU rental prices are rising, and compute can be asset-backed collateral, creating a long profitability flywheel.
Salesforce becomes AI company with Nvidia.
Salesforce is becoming an AI company by using Nvidia Neotron and agent technology to build proprietary agents on top of its data and application layers, while Nvidia accelerates data processing across Salesforce, Databricks, and Snowflake; this elevates Salesforce's ambition and growth opportunity.
Own Nvidia, don't trade it.
Nvidia was unfairly hit by fears that OpenAI won't IPO and won't make its quarter, but it trades at only about 14x FY28 earnings with enormous growth and dominant AI positioning; own it, don't trade it.
Salesforce is undervalued AI software winner.
Salesforce is an AI winner: it is adding agents and flexible outcome-based pricing, Slack is growing triple digits in bookings and is used by Nvidia, OpenAI, and Anthropic, the company bought back stock around 9x earnings and has a large Anthropic stake, guidance was raised, and management reaffirmed the 2030 goal with accelerating growth.
SaaS apocalypse is nonsense, software grows.
The SaaS apocalypse narrative is nonsense; AI is not killing enterprise software and instead can enhance SaaS products, as shown by continued Salesforce product growth and partnerships.
Bought Salesforce at low valuation.
Cramer bought Salesforce when it fell to about 9x earnings amid worries that Benioff had wrecked the balance sheet; he sees the roughly $25B Anthropic stake as a huge win that can cancel debt, and the stock still has growth optionality at around 15x earnings.
OpenAI still needs more compute.
Even if OpenAI paces frontier AI development for safety, it remains highly focused on getting more compute; more compute lets it build the best frontier models and cheaper child models, lowers prices, expands demand, and drives high returns on gigawatts, keeping the compute flywheel going.
Cybersecurity demand is a big deal.
Cybersecurity is a major growth area because OpenAI's Astra can find nearly every software vulnerability and Codex can patch at machine speed; the Hugging Face incident shows a closing defender window, so defenders need these tools today.
AI chip supply chain diversifies and grows.
OpenAI is diversifying its AI compute supply chain across multiple providers, including Nvidia for training, Broadcom for custom inference chips, AMD MI455 alternatives, TSMC wafers, and partners like Cerebras, to reduce risk and improve latency, reliability, and pricing.
Broadcom builds OpenAI inference chip.
Broadcom is a key OpenAI supplier because its custom Jalapeño inference chip is set up specifically for OpenAI models, making it important to OpenAI's inference buildout and supply-chain diversification.
AMD is viable OpenAI alternative.
AMD MI455s have become a very viable alternative for OpenAI in many cases, and OpenAI is working with partners like Cerebras as part of its supply-chain diversification, which can improve latency, reliability, and pricing.
This CNBC video, published September 15, 2026,
features Jim Cramer, Jensen Huang, Mark Benioff, Sarah Friar
discussing XLK, NVDA, CRM, SAAS, AI compute, CIBR, SMH, AVGO, AMD.
12 trade ideas extracted by AI with direction and confidence scoring.
Speakers:
Jim Cramer,
Jensen Huang,
Mark Benioff,
Sarah Friar
· Tickers:
XLK,
NVDA,
CRM,
SAAS,
AI compute,
CIBR,
SMH,
AVGO,
AMD
