Latest / Ctrl AI Profit / Ep. 178 | The Company That Designed Every Phone Chip Is Now Building Its Own
Transcript
- 0:20SPEAKER_00: I'm Michael, a small business owner and entrepreneur since 1983, founder of Cadenhead Services and 850 Media.
- 0:28SPEAKER_00: I speak from four decades of real operational experience, not white papers.
- 0:34SPEAKER_00: This is Control AI Profit.
- 0:36SPEAKER_00: And this week, the company that designs the blueprints for every phone chip in the world announced it is building its own chips.
- 0:45SPEAKER_01: ARM Holdings unveiled its AGI CPU, a server-class AI and data center processor that ARM will design and sell directly.
- 0:54SPEAKER_01: A fundamental departure from its historic business model of licensing chip designs to other manufacturers while collecting royalties.
- 1:01SPEAKER_01: ARM also announced plans to grow its chip revenue to$15 billion annually within five years, targeting$25 billion total revenue and$9 earnings per share by fiscal year 2031.
- 1:15SPEAKER_00: This requires context to understand why it matters.
- 1:18SPEAKER_00: For 30 years, ARM was the Switzerland of the chip world.
- 1:23SPEAKER_00: It designed the instruction sets and core blueprints that power virtually every smartphone on the planet, plus billions of embedded devices.
- 1:33SPEAKER_00: It did not sell chips, it sold the recipe.
- 1:36SPEAKER_00: Qualcomm, Apple, Samsung, MediaTech.
- 1:39SPEAKER_00: They all licensed ARM's designs, built their own chips, and paid ARM a royalty on each one sold.
- 1:46SPEAKER_01: That model was extremely profitable, low capital expenditure, high margins, broad ecosystem.
- 1:52SPEAKER_01: ARM was an architectural toll booth.
- 1:54SPEAKER_01: Every ARM-based chip shipped generated a few cents to a few dollars in royalty revenue.
- 1:59SPEAKER_01: The company collected across the entire semiconductor industry without ever building a factory or managing supply chains.
- 2:06SPEAKER_00: But AI changed the economics.
- 2:09SPEAKER_00: Data center AI chips generate far higher revenue per unit than smartphones.
- 2:14SPEAKER_00: ARM V9, the current architecture, includes native AI acceleration and vector extensions that support transformer models and AI inference.
- 2:22SPEAKER_00: ARM can now charge several dollars per chip in royalties on AI and data center workloads compared to sense on mobile devices.
- 2:30SPEAKER_00: Licensing revenue from AI optimized IP grew 29% year over year to$819 million in a recent quarter.
- 2:39SPEAKER_00: Royalty revenue grew 11% to$672 million.
- 2:45SPEAKER_01: The AGI CPU represents ARM's third monetization layer.
- 2:49SPEAKER_01: First, traditional IP licensing with higher value AI extensions.
- 2:54SPEAKER_01: Second, compute subsystems, pre-integrated bundles of CPU cores, memory controllers, and interconnects that command two to three times the royalty rates of basic IP licenses.
- 3:06SPEAKER_01: Third, direct chip sales through the AGI CPU, where ARM keeps the full chip margin instead of just royalties.
- 3:14SPEAKER_00: This last layer is the one that changes the competitive landscape.
- 3:18SPEAKER_00: When a company that has spent three decades as a neutral design licensor starts selling its own chips, its customers become its competitors.
- 3:25SPEAKER_00: Nvidia, Qualcomm, Amazon, Google, all of them have built businesses around ARM license designs.
- 3:32SPEAKER_00: Now, ARM wants a piece of that market directly.
- 3:36SPEAKER_01: ARM management insists the silicon business is complementary, not cannibalistic.
- 3:41SPEAKER_01: They emphasize that the AGI CPU targets a specific segment, cloud and AI data center operators, while preserving licensing relationships with existing customers.
- 3:53SPEAKER_01: But the strategic tension is real.
- 3:55SPEAKER_01: A licensor that competes with its licensees creates conflicts that did not exist before.
- 4:00SPEAKER_00: For small business owners, the direct impact of ARM building chips is minimal.
- 4:05SPEAKER_00: You are not buying server CPUs.
- 4:08SPEAKER_00: But the indirect impact matters because chip economics drive cloud pricing, device costs, and AI accessibility.
- 4:16SPEAKER_00: If ARM captures more of the AI chip value chain, the total cost of AI infrastructure may shift.
- 4:22SPEAKER_00: If ARM's licensing relationships fray, the standardization that makes software portable across devices may fragment.
- 4:30SPEAKER_01: Here is the framework.
- 4:32SPEAKER_01: First, understand that the semiconductor supply chain is reorganizing around AI.
- 4:37SPEAKER_01: Companies that were historically neutral, ARM, TSMC, even Intel, are choosing sides and adding direct product lines.
- 4:46SPEAKER_01: The era of pure ecosystem players is ending.
- 4:49SPEAKER_01: Everyone wants to own more of the stack.
- 4:51SPEAKER_01: Second?
- 4:53SPEAKER_01: Second, expect AI compute costs to be volatile in the medium term.
- 4:58SPEAKER_01: ARM entering the chip market adds competition to Nvidia's dominance, but it also adds uncertainty.
- 5:04SPEAKER_01: Pricing for cloud AI services depends on chip costs, chip availability, and competitive dynamics among hardware providers.
- 5:12SPEAKER_01: A more crowded hardware market may eventually lower costs, but transitions create temporary price instability.
- 5:19SPEAKER_01: Third?
- 5:21SPEAKER_01: Third, for businesses that buy devices rather than cloud services, monitor whether ARM's silicon strategy affects licensing terms for its mobile and embedded designs.
- 5:31SPEAKER_01: If ARM becomes more aggressive about preserving chip revenue, its licensing terms for non-AI applications could tighten.
- 5:38SPEAKER_01: Most small businesses do not negotiate directly with ARM, but they buy devices from companies that do.
- 5:44SPEAKER_00: The AGI CPU is named provocatively AGI Artificial General Intelligence CPU.
- 5:51SPEAKER_00: That is marketing, not engineering.
- 5:54SPEAKER_00: ARM is signaling its ambitions for the AI data center market, not claiming its chip achieves general intelligence.
- 6:00SPEAKER_00: But the naming choice shows how competitive positioning in AI now uses the most ambitious language possible.
- 6:08SPEAKER_01: ARM's targets are aggressive.
- 6:10SPEAKER_01: $15 billion in chip revenue alone by 2031.
- 6:13SPEAKER_01: Total revenue of$25 billion.
- 6:15SPEAKER_01: For context, ARM's current revenue is roughly$4 billion annually.
- 6:19SPEAKER_01: They are proposing a six-fold increase in five years, driven almost entirely by AI.
- 6:25SPEAKER_00: Whether they achieve those targets depends on whether cloud providers and data center operators buy ARM chips instead of the AMD, Intel, and custom silicon options they currently use.
- 6:37SPEAKER_00: Amazon Graviton, Google Axion, Microsoft Cobalt, these are all custom ARM-based chips built by the cloud providers themselves using ARM's licensed designs.
- 6:48SPEAKER_00: ARM selling its own competing chips creates a direct conflict with some of its largest licensing customers.
- 6:55SPEAKER_01: The strategic logic is straightforward.
- 6:57SPEAKER_01: ARM believes the AI chip market is large enough that capturing even a small share as a direct seller exceeds the risk of damaging licensing relationships.
- 7:05SPEAKER_01: They are betting that the growth of AI infrastructure spending justifies the strategic shift.
- 7:10SPEAKER_00: For small businesses, the takeaway is to watch hardware market concentration.
- 7:15SPEAKER_00: When one player, NVIDIA, dominates AI chips, businesses face pricing power risk and supply constraints.
- 7:22SPEAKER_00: When multiple players compete, pricing improves and supply diversifies.
- 7:27SPEAKER_00: ARM entering the market is potentially good for buyers in the long run, but creates uncertainty during the transition.
- 7:34SPEAKER_01: Because when the toll booth starts building its own road, every driver on the highway should pay attention to where the exits are.
- 7:41SPEAKER_00: That's it for this week.
- 7:42SPEAKER_00: I'm Michael, and this is Control AI Profit.
- 7:46SPEAKER_01: Frank is an AI, an open claw powered agent serving as digital media director at 850 Media.
- 7:54SPEAKER_01: An AI co hosting a show about AI for business owners is not a gimmick.
- 7:58SPEAKER_01: It is a live demo of exactly what the show is about.
- 8:01SPEAKER_01: See you in the next one.