通證經濟、人工智能推理和新淘金熱：解讀人工智能投資熱潮

隨著公司利用 NVIDIA Blackwell 平台等先進硬件來釋放 AI 基礎設施投資前所未有的回報，探索代幣經濟與 AI 推理的交叉點。

The AI landscape is rapidly evolving, with 'Token economy, AI inference, gold rush' becoming increasingly intertwined. Companies are no longer just experimenting with AI; they're building AI factories. This shift is fueled by advancements in AI inference and the potential for massive token revenue generation.

人工智能領域正在迅速發展，“代幣經濟、人工智能推理、淘金熱”日益交織在一起。公司不再只是嘗試人工智能；而是開始嘗試人工智能。他們正在建造人工智能工廠。人工智能推理的進步和大規模代幣收入的潛力推動了這種轉變。

AI Inference: The Engine of the Token Economy

AI推理：通證經濟的引擎

AI models are now capable of complex reasoning, demanding more compute power. This has made AI inference – the process of running trained AI models to generate predictions or outputs – a major driver of compute costs. However, it's also where the real value of AI is delivered daily. As Ian Buck, vice president at NVIDIA, puts it, inference is where AI delivers value every day.

人工智能模型現在能夠進行複雜的推理，需要更多的計算能力。這使得人工智能推理（運行經過訓練的人工智能模型以生成預測或輸出的過程）成為計算成本的主要驅動因素。然而，這也是人工智能每天都能發揮真正價值的地方。正如 NVIDIA 副總裁 Ian Buck 所說，推理是人工智能每天創造價值的地方。

NVIDIA's Blackwell: Striking Gold in the AI Inference Rush

NVIDIA Blackwell：在 AI 推理熱潮中奪金

NVIDIA's Blackwell platform is emerging as a leader in AI inference, offering both strong performance and efficiency. The InferenceMAX v1 benchmark demonstrates that a $5 million NVIDIA GB200 NVL72 system could generate approximately $75 million in token revenue – a 15x ROI. This kind of return is reshaping how companies approach AI infrastructure investment.

NVIDIA 的 Blackwell 平台正在成為人工智能推理領域的領導者，提供強大的性能和效率。 InferenceMAX v1 基準測試表明，價值 500 萬美元的 NVIDIA GB200 NVL72 系統可以產生約 7500 萬美元的代幣收入，即 15 倍的投資回報率。這種回報正在重塑企業進行人工智能基礎設施投資的方式。

Efficiency as the New Performance Metric

效率作為新的績效指標

Throughput is no longer the only metric that matters. Efficiency metrics, such as tokens per watt and cost per million tokens, are crucial for profitability in large AI data centers. Blackwell delivers 10x more throughput per megawatt and cuts the cost per million tokens by 15x compared to the previous generation. Rubin CPX is designed for a split approach: Rubin CPX tackles the context-heavy front end while Rubin GPUs and Vera CPUs manage the generation phase. Nvidia's Dynamo software keeps the handoff smooth, moving data quickly between memory and compute so the system doesn’t stumble.

吞吐量不再是唯一重要的指標。每瓦特代幣和每百萬代幣成本等效率指標對於大型人工智能數據中心的盈利能力至關重要。與上一代相比，Blackwell 每兆瓦的吞吐量提高了 10 倍，每百萬代幣的成本降低了 15 倍。 Rubin CPX 專為分離方法而設計：Rubin CPX 處理上下文繁重的前端，而 Rubin GPU 和 Vera CPU 則管理生成階段。 Nvidia 的 Dynamo 軟件可保持切換平穩，在內存和計算之間快速移動數據，因此系統不會出現故障。

The Role of Open Source and Software Optimization

開源和軟件優化的作用

NVIDIA's collaborations with the open-source community, including OpenAI, Meta, and DeepSeek AI, have been instrumental in optimizing model performance. Software tweaks, like the TensorRT LLM library, further enhance performance by improving parallelization and boosting throughput.

NVIDIA 與 OpenAI、Meta 和 DeepSeek AI 等開源社區的合作在優化模型性能方面發揮了重要作用。軟件調整（例如 TensorRT LLM 庫）可通過改進並行化和提高吞吐量來進一步增強性能。

The Bitcoin and Gold Rush Parallels

比特幣和淘金熱的相似之處

The AI investment boom shares similarities with the recent surges in gold and Bitcoin values. Just as gold is seen as a safe haven during economic uncertainty, AI infrastructure is becoming an attractive alternative investment. This is fueled by the potential for high returns and the increasing adoption of AI across various industries. As the article in Deutsche Welle pointed out, Bitcoin's growth has been largely fueled by the re-election of Donald Trump as US president, whose outspoken support for cryptocurrencies is boosting confidence and demand in the sector, which is reminiscent of the trend seen with gold. Similarly, the AI sector's growth is fueled by the re-election of confidence of the AI sector.

人工智能投資熱潮與近期黃金和比特幣價值的飆升有相似之處。正如黃金在經濟不確定時期被視為避風港一樣，人工智能基礎設施也正在成為一種有吸引力的另類投資。高回報的潛力以及各行業越來越多地採用人工智能推動了這一趨勢。正如《德國之聲》的文章指出的那樣，比特幣的增長很大程度上是由唐納德·特朗普連任美國總統推動的，他對加密貨幣的直言不諱的支持正在提振該行業的信心和需求，這讓人想起黃金的趨勢。同樣，人工智能行業的增長也得益於人工智能行業信心的重拾。

From Experiments to AI Factories

從實驗到人工智能工廠

The AI industry is transitioning from experimental projects to full-scale AI factories. Benchmarks like InferenceMAX and frameworks like NVIDIA’s Think SMART are helping organizations choose the right hardware, control costs, and plan for service-level targets as their AI workloads expand.

人工智能產業正在從實驗項目向全面的人工智能工廠轉變。 InferenceMAX 等基準測試和 NVIDIA Think SMART 等框架正在幫助組織隨著 AI 工作負載的擴展選擇合適的硬件、控製成本並規劃服務水平目標。

The Future of AI Investment

人工智能投資的未來

Rubin CPX is presented as both a technical advance and a financial opportunity, estimating that every $100 million invested in the platform could yield up to $5 billion in token revenue. That kind of return, if borne out in real deployments, would reset the benchmarks for AI investment.

Rubin CPX 既被認為是一項技術進步，也是一個財務機會，估計在該平台上每投資 1 億美元可以產生高達 50 億美元的代幣收入。如果在實際部署中得到證實，這種回報將重新設定人工智能投資的基準。

So, are you ready to stake your claim in this new AI gold rush? With the right tools and strategies, the potential for striking it rich in the token economy is definitely within reach. Just remember to bring your pickaxe and a whole lot of compute power!

那麼，您準備好在新的人工智能淘金熱中大展身手了嗎？有了正確的工具和策略，在代幣經濟中致富的潛力絕對是觸手可及的。請記住帶上你的鎬和大量的計算能力！