---
type: "Topics"
locale: "en"
url: "https://longbridge.com/en/topics/38937711.md"
description: "Try to understand Wall Street's cautionary logic towards AI.1. Large model companies are still operating at a loss, with no clear path to profitability in the short term.2. The money tech giants are spending on AI is excessively large, which could pose risks.3. If large model companies still cannot see a path to profitability in the future, and tech giants cannot effectively recoup their AI capital expenditures, it will be difficult to sustain such massive capital spending over the long term.4. That's why NVIDIA's current performance is dazzling, but its stock price remains unmoved.Solutions.5. The key still lies with semiconductor companies like NVIDIA to continuously reduce token costs.Assuming a large model company's revenue is 10 and its cost is 20, with token cost being 10, if the performance improvements from semiconductor chips like NVIDIA's can lower the token cost to 1, then profitability becomes a real possibility, and the entire business model outlined above becomes viable.6. The core is that large models must indeed keep improving, becoming capable enough for commercial production and operation, and effectively reducing costs for traditional companies.If the above two points can be achieved, then AI stocks will have a bright future 😁$NVIDIA(NVDA.US)$Alphabet - C(GOOG.US)$Taiwan Semiconductor(TSM.US)"
datetime: "2026-02-27T08:05:43.000Z"
locales:
  - [en](https://longbridge.com/en/topics/38937711.md)
  - [zh-CN](https://longbridge.com/zh-CN/topics/38937711.md)
  - [zh-HK](https://longbridge.com/zh-HK/topics/38937711.md)
author: "[躺平de交易员](https://longbridge.com/en/profiles/14754317.md)"
---

# Try to understand Wall Street's cautionary logic t…


### Related Stocks

- [SOXX.US](https://longbridge.com/en/quote/SOXX.US.md)
- [NVDA.US](https://longbridge.com/en/quote/NVDA.US.md)
- [SOXS.US](https://longbridge.com/en/quote/SOXS.US.md)
- [SOXL.US](https://longbridge.com/en/quote/SOXL.US.md)
- [CP00062.US](https://longbridge.com/en/quote/CP00062.US.md)

## Comments (10)

- **小宝不剩饭 · 2026-02-27T16:25:16.000Z**: So how can we effectively reduce token costs?
- **Seven. · 2026-02-27T09:52:08.000Z**: What should Dazi do now?
  - **躺平de交易员** (2026-02-27T10:03:32.000Z): I'm still lying flat.
- **Erstwhile · 2026-02-27T08:35:44.000Z**: The scaling effect of AI is beginning to show, expanding AI computing power has driven down the cost per token. To reduce costs, the current rapid expansion is a necessary stage to go through, and this part will be continuously amortized over time.
  - **躺平de交易员** (2026-02-27T08:40:23.000Z): Waiting for the future of the industry
  - **Erstwhile** (2026-02-27T08:49:23.000Z): I'm going to strangle the AI industry$Taiwan Semiconductor(TSM.US)🤣
  - **躺平de交易员** (2026-02-27T08:50:35.000Z): This is great, I want to buy it too 🤣
- **Kimui · 2026-02-27T08:11:11.000Z**: Just like the internet era, Cisco started as a shovel, but what comes next? Chips must drop in price, to a level that allows more companies to afford and use them, instead of constantly hyping shortages and price hikes (NVIDIA and its upstream suppliers). This is when the explosion of downstream app
  - **豐川祥子的交易员** (2026-02-27T08:20:43.000Z): Should we buy the upstream shovel or the downstream? It feels like the downstream is more cost-effective this way.
  - **躺平de交易员** (2026-02-27T08:22:09.000Z): Different people have different opinions😁
