Feature
eBook
Designing an AI-Native Contact Center
In another era, semiconductor chip manufacturing would be a great example of cooperation in global commerce. Chips get designed in the US and China, manufactured in Taiwan by the Taiwan Semiconductor Manufacturing Company (TSMC) using machinery largely created by Advanced Semiconductor Materials Lithography (ASML) in the Netherlands and then get shipped all over the world to be used in products.
Instead, countries are limiting which chips can be sold to other countries, slapping tariffs on chips and the equipment used to make them and threatening to take control of other countries’ facilities within their borders.
How did we get here?
Table of Contents
- Blame Artificial Intelligence
- The Role of Memory
- Europe’s Part in the AI Chip Supply Chain
- AI Customers at the Bottom of the Totem Pole
- How Enterprises Should Plan
Blame Artificial Intelligence
The problem is that AI has caused a huge demand for graphic processing units (GPUs) — or the chips that power AI — with the United States locked in a battle with China for dominance in AI.
While the US has invested billions in increasing its chip manufacturing facilities, starting with $50 billion for the CHIPS and Science Act of 2022, that carrot wasn’t enough. Instead, President Joe Biden also had to apply the stick: limits in January 2025 on which chips could be bought by which countries.
President Donald Trump has followed suit, limiting which advanced AI chips American companies such as NVIDIA were allowed to sell to countries such as China, as well as slapping tariffs up to 25% on chips. At the same time, Trump is also working with Taiwan to fund semiconductor manufacturing in the US, with Taiwan committing $250 billion toward the project.
But the chip global supply chain issue now goes beyond GPUs.
“The global semiconductor supply chain is deeply interdependent, and countries are working to protect their access to AI chips and hardware components that are critical for generative AI, high-performance computing and autonomous systems,” wrote Deloitte in a recent report.
“Therefore," the report continued, "it’s not surprising that export controls and other trade restrictions have started to affect a broader footprint of semiconductor equipment, materials, software, design tools, various kinds of chips and packaging and assembly tools in 2025 and 2026 compared to two or three years ago.”
The Role of Memory
Memory chips are becoming short in supply as well.
“This isn’t a baton-passing moment from GPUs to memory,” said Sanchit Vir Gogia, chief analyst, founder and CEO of Greyhound Research. “It’s a compounding problem. We’re not looking at a new bottleneck replacing an old one. We’re looking at a system where both constraints now coexist and feed off each other, turning AI infrastructure into a high-stakes balancing act.”
The problem is that powerful GPU-based systems are hitting ceilings not due to lack of cores, but lack of throughput, Gogia said. “The GPU is sitting there, technically present, but doing next to nothing because the system can’t move data fast enough in and out of memory.”
To make matters worse, when vendors do manufacture memory chips, they’re then stymied by a similar backup in packaging, which connects the memory to circuit boards. “The queue for that packaging is backed up,” Gogia said. “Even if you get the chip and the memory, you can’t ship anything if the packaging line is full.”
Related Article: Why the US and China Are Betting on Different AI Futures
Europe’s Part in the AI Chip Supply Chain
That said, the global AI chip supply chain isn’t simply a battle between the US and China. The European Union also has to play a role, in addition to that played by ASML.
The EU passed the European Chips Act in September 2023, intends to reduce the EU’s dependence on other countries. In September 2025, the 27 member states declared their intention for a Chips Act 2.0.
“Let us be clear: There is no AI continent without advanced chips,” said a spokesperson for the European Commission, the primary executive branch of the European Union. “With the Chips Act, we want to reach a global chips market share of 20% by 2030. Yes, this is ambitious but needed for our path to innovation, especially in making Europe a leader in AI."
With the Chips Act, the spokesperson added, the EU has already attracted more than €80 billion in investments. "85% of the total Chips for Europe Initiative budget has already been committed. This includes longer-term initiatives to support the development of AI chips in Europe.”
Other steps the Commission has taken to improve chip accessibility include:
- Holding a simulation exercise on semiconductor supply chain disruptions to improve Europe’s preparedness to respond to a crisis under the EU Chips Act
- Working with Taiwan, including investing €5 billion to support the European Semiconductor Manufacturing Co., a joint venture between Taiwan Semiconductor Manufacturing Co., Bosch, Infineon and NXP, in constructing and operating a microchip manufacturing plant in Dresden, Germany
- Supporting additional chip manufacturing facilities in Germany, Austria and Italy
AI Customers at the Bottom of the Totem Pole
So what does this all mean for the enterprise? They’re at the back of the line.
“NVIDIA, hyperscalers and a small group of AI-focused customers now sit at the center of the ecosystem and shape demand not just for chips, but for packaging, memory, substrates, power and networking,” said Brad Gastwirth, global head of research and market intelligence for Circular Technology.
“For end user companies, this means they sit downstream of a supply chain that prioritizes scale, long-term commitments and AI-specific builds. The impact shows up as longer lead times, tighter allocation and less flexibility than many companies were used to in prior cycles.”
In fact, that line is getting longer as companies in front invest even more heavily in chips.
“On average, hyperscalers went from a steady state CapEx of 12-15% of revenues in 2019-2022 to 25% in 2025,” said Jim Handy, general director of Objective Analysis. And behind them there’s more. “Nvidia appears to have convinced countries all over the world that they need to worry about data sovereignty, and that sentiment seems poised to perpetuate NVIDIA’s growth” in building regional data centers, such as the almost $1 trillion investment by Saudi Arabia to the US.
Related Article: 'Right-to-Compute' Laws May Be Coming to Your State This Year
Webinar
Feb
25
Content Strategy Leaders Live: Scaling for Speed, Complexity and AI in High Tech
A candid roundtable on how high-tech leaders are rethinking content at scale.
Register
Webinar
Feb
26
Do More with Less: Modernizing the Cloud Contact Center for 2026
Learn how to leverage cloud platforms without adding a single hire to personalize every customer interaction.
Register
Webinar
On demand
Cut the Noise: Deploying AI That Actually Moves the Needle
Learn how to turn AI experimentation into concrete revenue operations.
Watch Now
Webinar
On demand
Ditch the Desk Phones: How Modern Teams Drive AI-First Communications
Find out how one team finally pulled the plug on a legacy phone system. And built something smarter.
Watch Now
Webinar
On demand
Rebrand. Migrate. Optimize. How to Do It All (Without Slowing Down)
Cresta leveled up site speed, design flexibility and marketer sanity (in record time). Find out how.
Watch Now
Webinar
On demand
From Manual to Magical: How AI Transforms CX Teams
Learn how to replace manual support processes with automation that actually delivers.
Watch Now
How Enterprises Should Plan
This is not something that enterprises can hope will change, Gastwirth said.
“The market has moved from buyers choosing freely to capacity being assigned. Vendors that place large, multi-year orders naturally get priority in supply and roadmaps. For end users, regaining control does not mean reversing this dynamic. It means adapting to it.”
How do they do that? “AI infrastructure is no longer commodity IT,” according to Gastwirth. “It requires earlier planning, longer visibility and more flexibility. Practically, that means locking in requirements sooner, staying open to alternative system designs and building relationships beyond a single vendor. Many companies are also balancing cloud, on-prem and hybrid deployments so they are not dependent on one allocation path.”
But the result could be a stall in the enterprise projects AI companies need, Gogia said. “Systems designed a year ago assumed memory would be there when needed,” he said. “Now teams are revisiting decisions, downscaling model footprints and bolting on memory optimization tricks just to keep things running. You’ve got engineers designing around what’s in stock, not what’s ideal. That’s not a place you want to be when rolling out something mission-critical.”