📰 An AI Computing Power Revolution? Why Tsinghua and Microsoft’s 88% Cost Cut Could Change Everything?
Research teams from Tsinghua University and Microsoft used a new tool called HarnessMM to cut the token costs of DeepSeek AI by a whopping 88%. This means AI training and deployment could save a fortune in computing costs, potentially making large models affordable even for small companies and redefining the computational barriers across the entire AI industry.
Why does this news matter?
The core issue here is computing costs—the Achilles’ heel of AI. Training a large model can currently burn through hundreds of thousands or even millions of dollars, while the share price of computing giants like NVIDIA has become absurdly inflated. The Tsinghua-Microsoft partnership is no ordinary collaboration: Tsinghua is at the source of AI research, while Microsoft is a veteran in commercializing computing power. The fundamental reason they’re working together on this is to break NVIDIA’s monopoly on exorbitant profits and make AI truly accessible to everyone. With money continuing to flow out of tech stocks while the race to build large models remains fierce, this breakthrough in cutting costs and improving efficiency could mean explosive growth in AI adoption.
Impact on the market
In the short term, AI like DeepSeek, powered by cost-cutting and efficiency-boosting technology, could see a brief stock price surge. But in the medium to long term, this is more likely to shake up the pricing system for the entire AI infrastructure sector. If all AI companies can significantly reduce their computing costs, then:
1. BTC and ETH could see a new wave of computing-power narratives, as both crypto mining and AI applications become cheaper
2. Crypto projects still fixated on high computing-power barriers could be pushed to the sidelines
3. Regulators may start looking into how computing power is priced, which is more dangerous than simply focusing on token prices
A historical example is when OpenAI released GPT-3, driving demand for AI computing power across the board and sending NVIDIA’s share price soaring. But this time is different: open-source tools can directly cut costs, so the impact could be even greater.
💡 I believe a sharp drop in computing costs has more practical significance for ETH. ETH’s Layer 2 ecosystem already depends on low-cost computation. If the barrier to computing power comes down, use cases for ETH could spring up like mushrooms after rain. But this thesis would no longer hold if NVIDIA suddenly released a revolutionary, super-powerful chip.
This article is not sponsored by any project, and the author does not hold any of the assets mentioned.
⚠️ This is not investment advice. Predictions are for reference only.
$BTC $ETH
Research teams from Tsinghua University and Microsoft used a new tool called HarnessMM to cut the token costs of DeepSeek AI by a whopping 88%. This means AI training and deployment could save a fortune in computing costs, potentially making large models affordable even for small companies and redefining the computational barriers across the entire AI industry.
Why does this news matter?
The core issue here is computing costs—the Achilles’ heel of AI. Training a large model can currently burn through hundreds of thousands or even millions of dollars, while the share price of computing giants like NVIDIA has become absurdly inflated. The Tsinghua-Microsoft partnership is no ordinary collaboration: Tsinghua is at the source of AI research, while Microsoft is a veteran in commercializing computing power. The fundamental reason they’re working together on this is to break NVIDIA’s monopoly on exorbitant profits and make AI truly accessible to everyone. With money continuing to flow out of tech stocks while the race to build large models remains fierce, this breakthrough in cutting costs and improving efficiency could mean explosive growth in AI adoption.
Impact on the market
In the short term, AI like DeepSeek, powered by cost-cutting and efficiency-boosting technology, could see a brief stock price surge. But in the medium to long term, this is more likely to shake up the pricing system for the entire AI infrastructure sector. If all AI companies can significantly reduce their computing costs, then:
1. BTC and ETH could see a new wave of computing-power narratives, as both crypto mining and AI applications become cheaper
2. Crypto projects still fixated on high computing-power barriers could be pushed to the sidelines
3. Regulators may start looking into how computing power is priced, which is more dangerous than simply focusing on token prices
A historical example is when OpenAI released GPT-3, driving demand for AI computing power across the board and sending NVIDIA’s share price soaring. But this time is different: open-source tools can directly cut costs, so the impact could be even greater.
💡 I believe a sharp drop in computing costs has more practical significance for ETH. ETH’s Layer 2 ecosystem already depends on low-cost computation. If the barrier to computing power comes down, use cases for ETH could spring up like mushrooms after rain. But this thesis would no longer hold if NVIDIA suddenly released a revolutionary, super-powerful chip.
This article is not sponsored by any project, and the author does not hold any of the assets mentioned.
⚠️ This is not investment advice. Predictions are for reference only.
$BTC $ETH
