AI prices have plunged as much in 3 years as PC prices did in 15
Source: Barchart
Broadcom’s transformation is wild.
$AVGO quarterly revenue: 2019 → ~$6B Today → ~$30B VMware added a new engine. AI sent semiconductors soaring. Source: App Economy Insights
Memory prices have surged as AI datacenters demand a growing share of the world's DRAM supply.
Source: Hedgeye
Big tech giants booked a more than $160bn windfall last quarter from investments in other AI companies
Flattering their earnings and raising concerns that paper gains are overstating the strength of the AI boom. Source: FT
WE ARE EXPECTING AI PRODUCTIVITY GAINS FROM TECHNOLOGY MOST BUSINESSES AREN’T USING YET.
This chart can be read in two ways: AI companies have not yet built products that most businesses genuinely need. We are still at the very beginning of a massive new product category. Both are probably true. Today’s most commercially mature AI tools are coding agents. They can deliver real productivity gains—but mainly for developers and other technology-focused roles. The equivalent tools for finance, law, consulting, operations and administration are only starting to emerge. Cowork and Codex offer a glimpse of what is coming, but remain heavily oriented toward coding. So perhaps AI’s limited impact on productivity is not evidence that the technology has failed. It may simply reflect a more basic reality: We are trying to measure the economic impact of tools that most companies have not adopted yet. Source: Ramp, Ara Kharazian
Nomura has a chip shortage index
Source: Nomura
Tech has had quite the month so far... and we're not even halfway done
Source: Trend Spider
AI TOKEN COSTS ARE DOWN 40% FROM THEIR MAY PEAK.
The main reason? Chinese AI models are dramatically cheaper. DeepSeek, Qwen and Kimi are delivering increasingly competitive performance at 10–35x lower token prices than many U.S. alternatives — and already account for as much as 46% of usage in some enterprise segments. The pressure is spreading. OpenAI, Google and Anthropic have responded by cutting prices by as much as 80%. This creates an interesting dynamic: Cheaper inference → more AI adoption → more token consumption → more demand for compute. But at the same time, it is compressing margins and commoditizing the intelligence layer. The likely long-term winners could therefore be the hyperscalers and infrastructure providers: even if the price per token collapses, exploding volumes can more than compensate. AI intelligence may become a commodity. AI compute may not. And markets still seem far from fully pricing that shift. Source: Austral Research, zerohedge
