TSMC AI demand has outrun the world’s most advanced chipmaker’s ability to supply it, and the company’s CEO said on June 4 that the situation is not going to resolve anytime soon. Speaking at TSMC’s annual shareholders’ meeting in Taipei, C.C. Wei delivered the clearest public statement yet on the structural nature of the current shortage: global chip supply will continue to fall short of AI-driven demand for years, advanced nodes are oversold through at least 2027, and even TSMC’s $165 billion Arizona expansion will not be enough to fully satisfy American customers in the near term.
What TSMC’s CEO Said at the June 4 Shareholders’ Meeting
TSMC CEO C.C. Wei informed shareholders on June 4 that the global supply of chips will continue to lag behind AI-driven demand for years. TSMC’s first quarter 2026 revenue hit NT$1.13 trillion, roughly $35.7 billion, representing a 35% year-over-year increase driven almost entirely by the appetite for AI-capable silicon. The company projects more than 30% revenue growth for the full year 2026 in USD terms. Sound On Sound
Taiwan Semiconductor Manufacturing Co.’s global chip supply will fall short of AI-fueled demand for years to come, Wei said, suggesting production capacity remains a key bottleneck in the buildout of global computing infrastructure. TSMC won’t be able to fulfill demand led by American customers even as more manufacturing capacity comes online in the US over the next few years. Magnetic Magazine
Still, Taiwan’s largest company will refrain from initiating the sort of abrupt price hikes that shook up the memory chip sector. The decision to hold pricing steady while capacity is sold out is a deliberate strategic choice, designed to keep major clients like Nvidia, Broadcom, and AMD from seeking alternatives or accelerating their own foundry development programs. Magnetic Magazine
Advanced Node Capacity: Why Sub-5nm Chips Are 25-30% Undersupplied
Advanced-node capacity, the sub-7nm and sub-5nm chips that power everything from Nvidia’s latest GPUs to custom AI accelerators, has been under severe constraints since 2024. Demand at leading nodes is expected to exceed capacity by 25-30% in 2026, and the situation is not projected to ease until at least 2027. Sound On Sound
The numbers are striking. TSMC controls approximately 70% of global advanced semiconductor foundry capacity. If demand exceeds supply by 25-30% at the company that makes the majority of the world’s advanced chips, there is no realistic short-term path to equilibrium. The customers at the back of the queue, whether Nvidia competing for CoWoS packaging capacity or smaller AI accelerator startups trying to get their first wafers taped out, face multi-quarter or multi-year waits.
TSMC, as the world’s largest dedicated chip foundry with a market share of about 70% in 2025, is at the forefront of this crisis. Despite its efforts to expand capacity, the ongoing CoWoS shortage is expected to continue until at least 2027, complicating TSMC’s ability to meet the rising demand for AI computing chips. ManMade Music
Broadcom warns that TSMC capacity limits are creating AI chip bottlenecks through 2026, with PCB lead times surging to 6 months as supply chain strains intensify. GamesRadar+
TSMC Q1 2026 Revenue: $35.7B on 35% Year-Over-Year Growth
William Li, senior analyst at Counterpoint Research, told CNBC that AI chip demand has pushed TSMC’s manufacturing capacity to its limits. “The narrative for 2026 is as much about resource constraints as it is about growth. Demand still significantly outpaces supply and isn’t showing any major sign of slowing down,” Li said. “We expect this sold-out environment to remain a defining characteristic of the semiconductor industry throughout 2026, as semiconductor companies simply can’t keep products on their shelves,” he added. Tom’s Hardware
TSMC forecast full-year 2026 revenue growth of more than 30% year over year in U.S. dollar terms. Meanwhile, it projected second-quarter revenue of $39 billion to $40.2 billion, representing a 10% sequential increase. Tom’s Hardware
At its last earnings call in January, the company said it expected capital spending this year to rise as much as 37% to between $52 billion and $56 billion, reflecting its expansion efforts and an expectation that demand will remain strong. The company said it now expects capex to be at the high end of that range. Tom’s Hardware
This comes as the company faces concerns about supply chain disruptions linked to the Middle East conflict, including disruptions to energy supplies and key manufacturing materials such as helium and hydrogen. Tom’s Hardware
The $165 Billion Arizona Bet and Why It’s Already Behind Schedule
A $165 billion investment program is underway to expand production capacity in Arizona, where Wei said the two sites the company controls should cover its land needs through the mid-2030s. He cautioned, however, that a standing goal of concentrating at least 30% of leading-edge capacity below 2 nanometers on American soil is slipping out of reach, hampered by a slow environmental permitting process and an insufficient pool of skilled construction labor. It will take a “very long time” to fully satisfy American customers’ needs with U.S.-based production. Khel Now
The permitting bottleneck is a recurring story for semiconductor mega-projects in the United States. Environmental review processes that were designed for conventional industrial projects are poorly calibrated for the unique requirements of semiconductor fabs, which require extraordinary quantities of ultra-pure water, specialized chemicals, and extreme clean room conditions. Every month of permitting delay is a month of capacity that Nvidia, Apple, and AMD cannot access.
Wei described the Arizona site as a “gigafab cluster” aimed at boosting efficiency and helping serve US-based customers more effectively. The term “gigafab” refers to a manufacturing campus large enough to house multiple fabrication lines at different process nodes, reducing the logistical overhead of moving wafers between facilities and allowing TSMC to serve customers more efficiently at scale. esports.gg
Addressing labor compensation, Wei said bonuses tied to profit sharing have climbed by about 30% in each of the last two consecutive years, and a third straight year of similar gains is anticipated for 2026. Retaining the skilled workforce needed to operate leading-edge fabs in the United States requires compensation packages that can compete with what TSMC pays in Taiwan, where the cost of living is significantly lower. Khel Now
CoWoS Shortage: The Packaging Bottleneck No One Is Talking About
The headline shortage at TSMC is in wafer production. The less-discussed but equally consequential shortage is in chip-on-wafer-on-substrate (CoWoS) packaging, the advanced 2.5D packaging technology that allows multiple chips to be connected at high speed in a single package.
The ongoing CoWoS shortage is expected to continue until at least 2027. The supply bottlenecks in semiconductor manufacturing are particularly significant in the manufacturing of advanced 3nm to 2nm wafers and 2.5D/3D packaging. ManMade Music
CoWoS is what connects Nvidia’s H100 and H200 GPU dies to their high-bandwidth memory. Without it, even a manufactured GPU die cannot become a finished product. The bottleneck is not just silicon: it is the entire stack from wafer to packaged chip, and TSMC controls critical parts of every layer.
Optical transceiver PCB delivery schedules have ballooned from 6 weeks to half a year as the broader component shortage cascades through the supply chain beyond semiconductors alone. This cascading effect means that even companies which do not directly order from TSMC are feeling the shortage through their component suppliers. GamesRadar+
Will TSMC Raise Prices? C.C. Wei’s Answer at the Shareholders’ Meeting
Despite this supply-demand mismatch, Wei told shareholders the company would not resort to the aggressive price hikes seen in the memory chip sector, keeping major clients like Nvidia, Broadcom, and AMD from seeking alternatives. Sound On Sound
The memory chip sector reference is instructive. When DRAM and NAND flash memory manufacturers faced similar demand surges in previous cycles, some resorted to sharp price increases that damaged customer relationships and accelerated customers’ investments in alternative supply. TSMC’s decision to hold pricing steady while capacity is sold out reflects a longer-term view: the customers it needs for the next decade are more valuable than the short-term pricing power available today.
That discipline has a limit. The company faces concerns about supply chain disruptions linked to the Middle East conflict, including disruptions to energy supplies and key manufacturing materials such as helium and hydrogen. If input costs rise significantly, the no-price-hike commitment becomes harder to sustain. Tom’s Hardware
What the TSMC AI Demand Crunch Means for Nvidia, Broadcom, and AMD
The direct beneficiaries of TSMC’s capacity crunch are the companies that have already secured long-term supply agreements: primarily Nvidia, Apple, Broadcom, and AMD, in roughly that order of priority.
Founded in 1987, TSMC has established itself as a leader in the semiconductor industry, providing cutting-edge manufacturing services to major clients like Apple, AMD, and Nvidia. With a market capitalization of approximately $2.04 trillion, TSMC’s scale and technological prowess allow it to maintain solid operating margins, even in a highly competitive environment. ManMade Music
For Nvidia specifically, the capacity crunch is both the primary constraint on its revenue growth and the primary moat protecting its competitive position. If TSMC cannot make enough Blackwell chips to meet demand, Nvidia’s order book simply extends further into the future. Competitors cannot easily unseat Nvidia by offering better chips if TSMC is already allocating its capacity and cannot take new large customers for years.
This comes as Nvidia snaps up capacity as TSMC and Intel ramp chip packaging in the U.S. Tom’s Hardware
Intel as the Release Valve: Can It Fill the TSMC Capacity Gap?
TSMC’s production bottlenecks are shifting the semiconductor market landscape. One company gaining attention is Intel, which continues efforts to rebuild its own manufacturing capability after years of production delays. Intel does not need to match TSMC to benefit. Analysts suggest the company could serve as a release valve for an overloaded supply chain by offering foundry capacity to customers facing shortages. esports.gg
Intel’s facilities also offer geographic spread and greater alignment with US industrial priorities, factors that become more attractive as firms look to de-risk amid geopolitical tensions and trade uncertainty. Investor confidence in Intel appears to be strengthening. Shares have risen 19% in 2026 so far, buoyed by strong trading and greater belief in the company’s long-term plans. Momentum gained further when US President Donald Trump praised Intel’s latest processors after a meeting with CEO Lip-Bu Tan. esports.gg
Jake Lai, Senior Analyst at Counterpoint Research, tells CNBC that 2026 is shaping up to be a “breakout year” for AI servers, backed by growth in both chip manufacturing and packaging technology. esports.gg
Samsung’s foundry division and GlobalFoundries serve mature nodes but neither has the advanced node capability needed to absorb the overflow from TSMC’s most constrained products. The realistic near-term alternative is Intel Foundry Services, which is still ramping and carries execution risk but at least represents an option that did not exist three years ago.
Latest Updates
The shareholders’ meeting statements were made on June 4, 2026, in Taipei. CryptoBriefing confirmed that TSMC CEO C.C. Wei told shareholders that global chip supply will continue to lag AI demand for years, that Q1 2026 revenue hit $35.7 billion on 35% year-over-year growth, and that the company projects more than 30% full-year revenue growth while refusing to implement aggressive price hikes. CNBC confirmed from its April 2026 earnings coverage that Counterpoint Research analyst William Li said the sold-out environment would remain a defining characteristic of the semiconductor industry throughout 2026, with demand significantly outpacing supply. The Verge’s coverage of TSMC’s AI demand struggles provided the editorial framing that brought this story to a broader technology audience on June 4. Sound On SoundTom’s Hardware
Full sources: The Verge | CryptoBriefing | CNBC
Broader Implications
C.C. Wei’s shareholders’ meeting statement is the most direct public acknowledgment from any semiconductor executive that the AI chip shortage is structural rather than cyclical. It is not a supply chain disruption that clears in a few quarters. It is a fundamental mismatch between how fast AI compute demand is growing, doubling in token usage every three months by TSMC’s own data, and how fast humanity can build the fabs, train the workers, and permit the infrastructure needed to produce the chips.
The implications cascade in every direction. Cloud hyperscalers that cannot get enough GPUs delay their AI product launches. AI startups that cannot access compute run on shared infrastructure at higher cost. Countries that want domestic AI capability face a hardware bottleneck that no amount of software talent can overcome.
The $165 billion Arizona investment is the most visible response to that structural challenge, but Wei’s own admission that the permitting process and labor pool constraints are preventing it from meeting its capacity targets is a warning that even trillion-dollar commitments cannot overcome the mundane friction of industrial policy and labor markets.
For more semiconductor and AI infrastructure coverage, visit The Tech Marketer.
Frequently Asked Questions
1. What did TSMC’s CEO say about AI chip demand at the June 2026 shareholders’ meeting? TSMC CEO C.C. Wei told shareholders on June 4, 2026 that global chip supply will continue to lag AI-driven demand for years. He said advanced node capacity is expected to remain sold out through at least 2027, that the company projects more than 30% revenue growth for the full year in USD terms, and that TSMC will not resort to aggressive price hikes despite the demand-supply imbalance.
2. How undersupplied is TSMC’s advanced node capacity in 2026? Demand at TSMC’s leading nodes, sub-7nm and sub-5nm, is expected to exceed available capacity by 25 to 30% in 2026, according to analysis cited at the shareholders’ meeting. The CoWoS advanced packaging capacity that connects GPU dies to high-bandwidth memory is also in severe shortage, with relief not expected until at least 2027.
3. What is TSMC’s $165 billion Arizona investment and is it on schedule? TSMC is investing $165 billion in a semiconductor manufacturing campus in Arizona designed to serve US-based customers. Wei acknowledged at the June 4 shareholders’ meeting that a goal of concentrating at least 30% of leading-edge sub-2nm capacity in the US is slipping behind schedule due to slow environmental permitting and an insufficient pool of skilled construction labor.
4. Will TSMC raise chip prices in 2026? Wei explicitly told shareholders that TSMC will not resort to the aggressive price hikes that characterized the memory chip sector in previous demand surges. The company views stable pricing as essential to retaining its major customers, including Nvidia, Apple, Broadcom, and AMD, through a multi-year capacity expansion cycle.
5. How does the TSMC chip shortage affect companies like Nvidia and Intel? Nvidia’s revenue growth is directly constrained by how many Blackwell chips TSMC can produce. The shortage also represents a competitive moat: rivals cannot easily challenge Nvidia if TSMC has no capacity to manufacture their chips at scale. Intel, which is rebuilding its own foundry capability, is being watched as a potential overflow release valve for customers that cannot get TSMC allocation.
