There's been a hidden trend in the global tech scene recently: AI isn't just busy writing papers, drawing pictures, and generating fake videos—it's also driving the semiconductor industry to new heights. In short, wherever AI is booming, chip manufacturers are working overtime.

Here are some hard numbers to consider:

In the second quarter of this year, global spending on AI infrastructure skyrocketed to $82 billion. Yes, you read that right.

Among them, AI servers alone accounted for $80.4 billion, a staggering increase of over 2.7 times year-on-year!

Even more astonishing is that accelerated servers like GPUs and XPUs became the absolute mainstay, making up 91.8% of total AI server spending, with expenditures soaring to $73.8 billion—a nearly 3.1-fold increase compared to the previous year.

In comparison, spending on AI storage devices appears relatively "low-key," totaling only $1.6 billion, a year-on-year increase of 16.1%.

IDC predicts that by 2029, total spending on AI servers and storage will reach 758billion,withthestorageportionexpectedtobejust10.5 billion, accounting for about 1.4%. Computing power is truly the star of the show.

Regionally, the US has invested 63billioninAIinfrastructure,6.6timesthatofChina(whichspent9.5 billion in the same period).

Although the US is currently leading this wave of AI infrastructure investment, the Chinese market is also steadily growing, proving to be a real potential powerhouse.

The explosive growth of AI computing power has directly fueled the entire semiconductor market.

According to data from the U.S. Semiconductor Industry Association, global semiconductor sales surged to $208.4 billion in the third quarter of this year, representing a 15.8% increase quarter-over-quarter.

In September alone, sales reached $69.5 billion, marking a 25.1% increase compared to the same period last year.

Multiple categories, including memory chips and logic chips, are rebounding, with the Asia-Pacific and Americas regions serving as the primary growth engines.

On the technological front, semiconductors are also experiencing a wave of "mass production breakthroughs":

TSMC's 2nm process is expected to enter mass production later this quarter, with the A16 process also scheduled for production in the second half of the year;

HBM4 memory is about to hit the market, with SK Hynix planning mass production in Q4, and Samsung also planning to launch a 12-layer HBM4 product, with an expected price increase of over 60% compared to the current HBM3e;

Advanced packaging capacity is also expanding continuously. TSMC is actively increasing its CoWoS capacity, with plans to enlarge the reticle size by 5.5 times to accommodate more HBM4.

Additionally, Intel's Panther Lake processor, based on the 18A process, is expected to launch in early 2026, further boosting AI PC and server computing power. Autonomous driving chips, silicon photonics interconnects, RISC-V architecture, and silicon carbide substrates are also seeing concentrated deployment this year.

What's even more interesting is that AI and chips have entered a mode of "mutual empowerment":

AI is helping optimize chip design, and TSMC is expected to incorporate NVIDIA's cuLitho computational lithography platform in its 2nm R&D, using generative AI to accelerate process development—this is what you might call "you help me build chips, I help you get smarter."

It's foreseeable that as various advanced semiconductor technologies gradually enter mass production, AI infrastructure will move into a new phase of higher efficiency and greater consolidation. WSTS predicts global semiconductor sales will approach $700 billion in 2025, an 11.2% year-on-year increase.

The raging fire of AI appears to have first enriched the giants at the top of the chip supply chain—Nvidia, TSMC, SK Hynix... Their orders and production capacities have been driven to new heights by the global thirst for computing power.

But looking deeper, AI and semiconductors are forming a mutually reinforcing flywheel. Computing power, like water and electricity, is becoming a fundamental resource for the future world, while semiconductors are the "pipes" delivering this resource. The U.S. is betting heavily to solidify its technological hegemony, while China is steadily catching up, vying for a ticket to the next wave of industrial upgrades.

Behind these staggering numbers, the real answer may be this: The fire of AI will ultimately "enrich" the nations and companies that can independently control the cornerstone of computing power. Because this race is not just about who spends more money—it’s about who lays the groundwork earlier, who has stronger technology, and who builds a more complete ecosystem.

Semiconductors are the hardest foundation beneath AI’s feet, and only those who can build this foundation well will have the right to define the next decade.