<p>半导体行业在人工智能（AI）技术的发展中发挥了关键作用。从训练流行的 AI 模型到在数据中心运行推理应用，再到为智能手机、个人电脑和汽车等边缘设备提供 AI 工作负载，半导体通过快速执行大量计算来承担重任。</p>
<p>不同类型的芯片正在处理上述应用中的 AI 工作负载。这些芯片包括中央处理器（CPU）、图形处理器（GPU）和被称为应用特定集成电路（ASIC）的专用芯片。设计这些芯片的公司，如 <strong>Nvidia</strong>、<strong>Advanced Micro Devices</strong> 和 <strong>Broadcom</strong>，近年来增长加速显著。</p>
<p><img src="https://imageproxy.pbkrs.com/https://g.foolcdn.com/image//query-b3A9cmVzaXplJnVybD1odHRwczovL2cuZm9vbGNkbi5jb20vZWRpdG9yaWFsL2ltYWdlcy84MzMwMzgvYWktd3JpdHRlbi1vbi1jaXJjdWl0LWJvYXJkLmpwZyZ3PTcwMA?x-oss-process=image/auto-orient,1/interlace,1/resize,w_1440,h_1440/quality,q_95/format,jpg" alt="" width="700" height="376" original-src="https://imageproxy.pbkrs.com/https://g.foolcdn.com/image//query-b3A9cmVzaXplJnVybD1odHRwczovL2cuZm9vbGNkbi5jb20vZWRpdG9yaWFsL2ltYWdlcy84MzMwMzgvYWktd3JpdHRlbi1vbi1jaXJjdWl0LWJvYXJkLmpwZyZ3PTcwMA"/></p>
<p>图片来源：Getty Images。</p>
<p>然而，重要的是要注意，以上提到的所有公司都是无工厂芯片制造商，这意味着它们只设计芯片，而不进行生产。生产由它们的代工合作伙伴——<strong>台积电</strong>（TSM 2.20%）完成。作为一家总部位于台湾的代工巨头，台积电在这一技术的传播中发挥了开创性作用。</p>
<p>让我们看看为什么台积电可能是最重要的半导体公司，并且有潜力在本十年末重新定义 AI 行业。</p>
<h2>台积电的先进工艺节点在 AI 扩展中发挥关键作用</h2>
<p>台积电的制造厂使用不同的工艺节点生产芯片。先进的工艺节点往往能在单个芯片中集成更多的晶体管。因此，这些芯片可以同时提供更强的计算能力并降低能耗。</p>
<p>Nvidia 的 A100 数据中心 GPU 用于训练 ChatGPT，采用的是台积电的 7 纳米（nm）工艺节点。Nvidia 目前据说正在使用台积电的 4nm 工艺节点来制造其最新一代的 Blackwell AI GPU。这些新 GPU 在性能和能效方面相较于基于 5nm 工艺的前一代芯片有显著提升。</p>
<p>因此，台积电使 Nvidia 能够通过转向更好的工艺节点来提升其最新处理器的性能。因此，使用台积电先进工艺节点设计芯片的公司可以训练更大的模型，并在数据中心更高效地运行 AI 推理应用。在 AI 模型变得越来越高效、对 AI 数据中心能耗的担忧日益加剧的情况下，台积电制造的先进计算芯片的需求将会增加。</p>
<p>这正是台积电能够持续增加其全球代工市场份额的原因。它在第二季度控制了令人印象深刻的 70.2% 的全球代工市场份额，比第一季度增加了 260 个基点。</p>
<p>第二大代工厂商三星的市场份额仅为 7.3%。这并不令人惊讶，因为韩国公司的工艺节点在性能上据说不如台积电。但台积电并没有停止推动技术的边界。</p>
<h2>更先进的工艺节点即将到来</h2>
<p>台积电的 3nm 节点被 <strong>Apple</strong>、<strong>Qualcomm</strong> 和其他芯片设计公司广泛使用。预计该公司将在今年开始生产更先进的 2nm 工艺节点的芯片，承诺在晶体管密度上比 3nm 节点提高 15%。因此，台积电的 2nm 芯片可能在功耗上减少 25% 到 30%，并在性能上提高 10% 到 15%。</p>
<p>但这并不是公司要停止的地方。台积电管理层估计，到 2026 年底，它可以转向 1.6nm 芯片节点，随后在 2028 年转向 1.4nm 节点。该公司声称，1.4nm 工艺 “将在相同功率下提供高达 15% 的速度提升，或在相同速度下减少高达 30% 的功耗，同时逻辑密度提高超过 20%”，与 2nm 工艺相比。</p>
<p>最后，台积电估计到 2030 年可以转向更小的 1nm 节点。这意味着台积电的客户未来可以制造出更强大的芯片，同时降低功耗。这理想上将使台积电在长期内增加其全球代工市场份额，为公司的未来增长奠定基础。</p>
<p>毫不奇怪，分析师的共识估计预测台积电的盈利增长将加速。</p>
<p><img src="https://imageproxy.pbkrs.com/https://media.ycharts.com/charts/a512127ed048065728c4c1202b39db5d.png?x-oss-process=image/auto-orient,1/interlace,1/resize,w_1440,h_1440/quality,q_95/format,jpg" alt="" width="720" height="380" original-src="https://imageproxy.pbkrs.com/https://media.ycharts.com/charts/a512127ed048065728c4c1202b39db5d.png"/></p>
<p>数据来源：YCharts。</p>
<p>凭借上述产品路线图，它可以在本十年末维持这一健康的增长率。假设台积电在 2029 年和 2030 年能够以 20% 的年增长率增加其盈利，到 2030 年其每股收益可能达到 19.38 美元（从 2028 年的 13.46 美元）。将其乘以以科技股为主的 <strong>纳斯达克 100</strong> 指数的前瞻性市盈率 26 倍，预计五年后股价将达到 511 美元，较当前水平潜在上涨 92%。</p>
<p>考虑到台积电的股票目前仅以 22 倍的前瞻性市盈率交易，投资者在这只半导体股票上获得了很好的交易。考虑到其吸引人的估值，投资者应立即考虑购买台积电的股票，因为公司在未来五年内通过更先进的芯片重新定义行业的能力可能导致超出预期的盈利增长，这可能导致其股价比上述预测更大幅度上涨。</p>

台积电

英伟达

半导体 3 倍做多 - Direxion

<p>半导体行业对人工智能技术至关重要，像英伟达和 AMD 这样的公司依赖台积电进行芯片制造。台积电的先进工艺节点，如 7nm 和 4nm，提升了人工智能应用的性能和效率。台积电在全球代工市场占有 70.2% 的份额，并计划在 2028 年前推出 2nm 和 1.4nm 节点，承诺在功耗和性能方面实现显著改善。分析师预测，台积电的收益可能每年增长 20%，到 2030 年其股价可能上涨 92%</p>

<p>The semiconductor industry has played a pivotal role in the growth of artificial intelligence (AI) technology. From training popular AI models to running inference applications in data centers to powering AI workloads on edge devices such as smartphones, personal computers, and cars, semiconductors do the heavy lifting by performing large amounts of calculations quickly.</p>
<div class="lb-trans"><p>半导体行业在人工智能（AI）技术的发展中发挥了关键作用。从训练流行的 AI 模型到在数据中心运行推理应用，再到为智能手机、个人电脑和汽车等边缘设备提供 AI 工作负载，半导体通过快速执行大量计算来承担重任。</p>
</div><p>Different types of chips are tackling AI workloads in the applications mentioned above. They include central processing units (CPUs), graphics processing units (GPUs), and specialized chips known as application-specific integrated circuits (ASICs). Companies that design these chips, such as <strong>Nvidia</strong>, <strong>Advanced Micro Devices</strong>, and <strong>Broadcom</strong>, have seen solid acceleration in growth in recent years.</p>
<div class="lb-trans"><p>不同类型的芯片正在处理上述应用中的 AI 工作负载。这些芯片包括中央处理器（CPU）、图形处理器（GPU）和被称为应用特定集成电路（ASIC）的专用芯片。设计这些芯片的公司，如 <strong>Nvidia</strong>、<strong>Advanced Micro Devices</strong> 和 <strong>Broadcom</strong>，近年来增长加速显著。</p>
</div><p><img src="https://imageproxy.pbkrs.com/https://g.foolcdn.com/image//query-b3A9cmVzaXplJnVybD1odHRwczovL2cuZm9vbGNkbi5jb20vZWRpdG9yaWFsL2ltYWdlcy84MzMwMzgvYWktd3JpdHRlbi1vbi1jaXJjdWl0LWJvYXJkLmpwZyZ3PTcwMA" alt="" original-src="https://g.foolcdn.com/image/?url=https%3A%2F%2Fg.foolcdn.com%2Feditorial%2Fimages%2F833038%2Fai-written-on-circuit-board.jpg&amp;op=resize&amp;w=700"/></p>
<p>Image source: Getty Images.</p>
<div class="lb-trans"><p>图片来源：Getty Images。</p>
</div><p>It's important to note, however, that all the above-mentioned names are fabless chipmakers, which means that they only design the chips and don't manufacture them. The manufacturing is done by their foundry partner -- <strong>Taiwan Semiconductor Manufacturing</strong>(TSM 2.20%). Popularly known as TSMC, this Taiwan-based foundry giant has played a pioneering role in the spread of this technology.</p>
<div class="lb-trans"><p>然而，重要的是要注意，以上提到的所有公司都是无工厂芯片制造商，这意味着它们只设计芯片，而不进行生产。生产由它们的代工合作伙伴——<strong>台积电</strong>（TSM 2.20%）完成。作为一家总部位于台湾的代工巨头，台积电在这一技术的传播中发挥了开创性作用。</p>
</div><p>Let's see why TSMC is probably the most important semiconductor company out there and has the potential to redefine the AI industry by the end of the decade.</p>
<div class="lb-trans"><p>让我们看看为什么台积电可能是最重要的半导体公司，并且有潜力在本十年末重新定义 AI 行业。</p>
</div><h2>TSMC's advanced nodes play a key role in the expansion of AI</h2>
<div class="lb-trans"><h2>台积电的先进工艺节点在 AI 扩展中发挥关键作用</h2>
</div><p>TSMC's fabrication plants produce chips using different process nodes. Advanced process nodes tend to pack more transistors into a single chip. As a result, these chips can deliver more computing power and reduce energy consumption simultaneously.</p>
<div class="lb-trans"><p>台积电的制造厂使用不同的工艺节点生产芯片。先进的工艺节点往往能在单个芯片中集成更多的晶体管。因此，这些芯片可以同时提供更强的计算能力并降低能耗。</p>
</div><p>Nvidia's A100 data center GPU, which was used for training ChatGPT, was manufactured on TSMC's 7-nanometer (nm) process node. Nvidia is now reportedly using TSMC's 4nm process node to manufacture its latest generation of Blackwell AI GPUs. These new GPUs deliver a significant jump in performance and power efficiency over the previous generation of chips that were based on a 5nm process.</p>
<div class="lb-trans"><p>Nvidia 的 A100 数据中心 GPU 用于训练 ChatGPT，采用的是台积电的 7 纳米（nm）工艺节点。Nvidia 目前据说正在使用台积电的 4nm 工艺节点来制造其最新一代的 Blackwell AI GPU。这些新 GPU 在性能和能效方面相较于基于 5nm 工艺的前一代芯片有显著提升。</p>
</div><p>So, TSMC enabled Nvidia to enhance the performance of its latest processors by moving to a better process node. As a result, companies designing chips using TSMC's advanced process nodes can train bigger models and run AI inference applications more efficiently in data centers. At a time when AI models are becoming more efficient and concerns are being raised about the amount of power consumed by AI data centers, the need for advanced computing chips that TSMC makes is going to increase.</p>
<div class="lb-trans"><p>因此，台积电使 Nvidia 能够通过转向更好的工艺节点来提升其最新处理器的性能。因此，使用台积电先进工艺节点设计芯片的公司可以训练更大的模型，并在数据中心更高效地运行 AI 推理应用。在 AI 模型变得越来越高效、对 AI 数据中心能耗的担忧日益加剧的情况下，台积电制造的先进计算芯片的需求将会增加。</p>
</div><p>This is precisely the reason TSMC has been able to consistently increase its share of the global foundry market. It controlled an impressive 70.2% of the global foundry market in the second quarter, up by 260 basis points from Q1.</p>
<div class="lb-trans"><p>这正是台积电能够持续增加其全球代工市场份额的原因。它在第二季度控制了令人印象深刻的 70.2% 的全球代工市场份额，比第一季度增加了 260 个基点。</p>
</div><p>The second-place foundry player, Samsung, has a market share of just 7.3%. That's not surprising as the Korean company's process nodes are reportedly inferior to TSMC's in terms of performance. But TSMC isn't done pushing the envelope yet.</p>
<div class="lb-trans"><p>第二大代工厂商三星的市场份额仅为 7.3%。这并不令人惊讶，因为韩国公司的工艺节点在性能上据说不如台积电。但台积电并没有停止推动技术的边界。</p>
</div><h2>More advanced process nodes are on the way</h2>
<div class="lb-trans"><h2>更先进的工艺节点即将到来</h2>
</div><p>TSMC's 3nm nodes are being widely used by <strong>Apple</strong>, <strong>Qualcomm</strong>, and other chip designers. The company is expected to start producing chips on an even better 2nm process node this year, promising a 15% improvement in transistor density over the 3nm node. As a result, TSMC's 2nm chips could reduce power consumption by 25% to 30% and enhance performance by 10% to 15% over the prior-generation chips.</p>
<div class="lb-trans"><p>台积电的 3nm 节点被 <strong>Apple</strong>、<strong>Qualcomm</strong> 和其他芯片设计公司广泛使用。预计该公司将在今年开始生产更先进的 2nm 工艺节点的芯片，承诺在晶体管密度上比 3nm 节点提高 15%。因此，台积电的 2nm 芯片可能在功耗上减少 25% 到 30%，并在性能上提高 10% 到 15%。</p>
</div><p>But that's not where the company is going to stop. TSMC management estimates that it can move to a 1.6nm chip node by the end of 2026, followed by a 1.4nm node in 2028. The company claims that the 1.4nm process &#34;will offer up to a 15% speed improvement at the same power, or up to a 30% power reduction at the same speed, along with more than a 20% increase in logic density&#34; as compared to the 2nm process.</p>
<div class="lb-trans"><p>但这并不是公司要停止的地方。台积电管理层估计，到 2026 年底，它可以转向 1.6nm 芯片节点，随后在 2028 年转向 1.4nm 节点。该公司声称，1.4nm 工艺 “将在相同功率下提供高达 15% 的速度提升，或在相同速度下减少高达 30% 的功耗，同时逻辑密度提高超过 20%”，与 2nm 工艺相比。</p>
</div><p>And finally, TSMC estimates that it can move to an even smaller 1nm node by 2030. What this means is that TSMC's customers can make even more powerful chips in the future with lower power consumption. This should ideally enable TSMC to increase its share of the global foundry market in the long run, setting the company up for outstanding growth going forward.</p>
<div class="lb-trans"><p>最后，台积电估计到 2030 年可以转向更小的 1nm 节点。这意味着台积电的客户未来可以制造出更强大的芯片，同时降低功耗。这理想上将使台积电在长期内增加其全球代工市场份额，为公司的未来增长奠定基础。</p>
</div><p>Not surprisingly, consensus analyst estimates project TSMC's earnings growth will accelerate.</p>
<div class="lb-trans"><p>毫不奇怪，分析师的共识估计预测台积电的盈利增长将加速。</p>
</div><p><img src="https://imageproxy.pbkrs.com/https://media.ycharts.com/charts/a512127ed048065728c4c1202b39db5d.png" alt="" original-src="https://media.ycharts.com/charts/a512127ed048065728c4c1202b39db5d.png"/></p>
<p>Data by YCharts.</p>
<div class="lb-trans"><p>数据来源：YCharts。</p>
</div><p>It can sustain this healthy growth rate through the end of the decade thanks to its product roadmap discussed above. Assuming TSMC can increase its earnings at an annual rate of 20% in 2029 and 2030, its bottom line could hit $19.38 per share in 2030 (from $13.46 in 2028). Multiplying that by the tech-heavy <strong>Nasdaq-100</strong> index's forward earnings multiple of 26 points toward a stock price of $511 after five years, a potential jump of 92% from current levels.</p>
<div class="lb-trans"><p>凭借上述产品路线图，它可以在本十年末维持这一健康的增长率。假设台积电在 2029 年和 2030 年能够以 20% 的年增长率增加其盈利，到 2030 年其每股收益可能达到 19.38 美元（从 2028 年的 13.46 美元）。将其乘以以科技股为主的 <strong>纳斯达克 100</strong> 指数的前瞻性市盈率 26 倍，预计五年后股价将达到 511 美元，较当前水平潜在上涨 92%。</p>
</div><p>Given that TSMC stock trades at just 22 times forward earnings right now, investors are getting a great deal on this semiconductor stock. They should consider buying TSMC stock right away, considering its attractive valuation, as the company's ability to redefine the industry with even more advanced chips over the next five years could lead to stronger-than-expected earnings growth, and that could result in a sharper jump in its stock price than predicted above.</p>
<div class="lb-trans"><p>考虑到台积电的股票目前仅以 22 倍的前瞻性市盈率交易，投资者在这只半导体股票上获得了很好的交易。考虑到其吸引人的估值，投资者应立即考虑购买台积电的股票，因为公司在未来五年内通过更先进的芯片重新定义行业的能力可能导致超出预期的盈利增长，这可能导致其股价比上述预测更大幅度上涨。</p>
</div>

Prediction: This AI Company Will Redefine Semiconductors by 2030

- 半导体行业在人工智能技术增长中发挥了关键作用。  
- TSMC 控制了全球代工市场 70.2% 的份额，推动了芯片性能提升。  
- 预计 TSMC 未来将实现更高的收益增长，股价有望大幅上涨。

预测：这家人工智能公司将在 2030 年重新定义半导体行业