Microsoft and Gen AI Adoption
Microsoft (MSFT): Satya Nadella’s hype hasn’t exactly driven mass adoption of Copilot Pro. If it weren’t for OpenAI, I’m not sure that corporate clients would choose Azure over AWS or GCP. MSFT had a less than stellar September quarter and gave a less than robust December quarter outlook. Other than price increases that kicked in during the December quarter, I’m not sure that MSFT has anything to write home about. Yet, MSFT shares have recaptured most of what they lost on October 31st after the earnings release (MSFT shares were off 6% from the previous close). Seems like a setup for disappointment.
The big winners in terms of Gen AI-related revenue are Nvidia (NVDA) and OpenAI. That’s it. Anthropic and Google (GOOGL) get honorable mention. In terms of what is holding LLM adoption back, there are two core categories:
Trust and Reliability: trust and reliability are two of the primary hesitations users have that are preventing wider adoption of LLMs and LLM-powered apps such as Copilot Pro. This stems from a lack of trust in LLM outputs caused by hallucinations (incorrect or nonsensical outputs, particularly in complex or ambiguous contexts) and other inconsistencies, such as the same prompt yielding varying outputs.
Uncertainty in ROI: users may struggle to quantify how much time or effort an LLM actually saves, particularly if they need to verify outputs manually.
Further, the monolithic frontier LLM build cycle won’t last forever as capital will 100% become a constraint. Capital requirements may reel in LLM development before the scaling laws do, unless scaling has already slowed as many believe.