Latest / DX Today | No-Hype Podcast & News About AI & DX / AI Data Centers Are Killing the $100 Smartphone

Transcript

AI Data Centers Are Killing the $100 Smartphone · DX Today | No-Hype Podcast & News About AI & DX

• 0:00
  SPEAKER_00: Welcome to the DX Today Podcast, your daily deep dive into the AI ecosystem.
• 0:04
  SPEAKER_00: I'm Chris, and joining me as always is Laura.
• 0:07
  SPEAKER_01: Hi, Chris, and hello everyone.
• 0:09
  SPEAKER_01: Today I want to talk about a phone.
• 0:11
  SPEAKER_01: Not a fancy flagship, but the humble hundred dollar smartphone, because it is quietly disappearing, and AI is a big reason why.
• 0:19
  SPEAKER_00: That is a surprising place to start, because when most of us think about AI, we picture giant models and chatbots.
• 0:26
  SPEAKER_00: Not the cheap handset someone buys at a market stall in Lagos or Jakarta.
• 0:29
  SPEAKER_01: Exactly.
• 0:30
  SPEAKER_01: And that is what makes this story so fascinating.
• 0:33
  SPEAKER_01: Shipments of smartphones that cost under$100 fell about 60% year over year in the second quarter of this year, according to reporting from Rest of World.
• 0:42
  SPEAKER_00: 60% is not a dip, that is a collapse.
• 0:45
  SPEAKER_00: Give me some sense of scale here, because I honestly have no idea how many of these budget phones are normally sold in a year.
• 0:52
  SPEAKER_01: Last year, roughly 173 million phones under$100 shipped worldwide.
• 0:58
  SPEAKER_01: And back then, more than one in four smartphones on the planet cost less than$150.
• 1:03
  SPEAKER_01: So this is a huge slice of the market.
• 1:06
  SPEAKER_00: Okay, so this is a massive category that is shrinking incredibly fast.
• 1:10
  SPEAKER_00: Walk me through the connection because I need to understand how a data center in Virginia or Texas ends up changing what a phone costs in India.
• 1:18
  SPEAKER_01: It all comes down to memory chips.
• 1:20
  SPEAKER_01: Every phone needs DRAM, the working memory, and NAND storage.
• 1:24
  SPEAKER_01: And the exact same handful of factories that make phone memory also make the high bandwidth memory that AI accelerators devour.
• 1:31
  SPEAKER_00: And I'm guessing there are not very many of those factories, which is usually where these supply chain stories start to get really uncomfortable for everybody downstream.
• 1:39
  SPEAKER_01: You guessed right.
• 1:40
  SPEAKER_01: Three companies, Samsung, SK Heinex, and Micron, control more than 90% of the global memory market.
• 1:47
  SPEAKER_01: Ramon Lamas at IDC told Rest of World that in late 2025, all three pivoted most of their supply toward AI.
• 1:55
  SPEAKER_00: Let me play devil's advocate for a second.
• 1:57
  SPEAKER_00: Companies chase the best margins all the time.
• 2:00
  SPEAKER_00: Why would phone memory become collateral damage instead of just getting a little more expensive at the edges?
• 2:05
  SPEAKER_01: Because the economics are brutal.
• 2:07
  SPEAKER_01: High bandwidth memory earns roughly three to five times more revenue per wafer than standard mobile memory.
• 2:13
  SPEAKER_01: And according to Micron, one wafer of high bandwidth memory eats the factory capacity that could have produced about three wafers of ordinary DDR5.
• 2:22
  SPEAKER_00: So that every wafer they shift toward AI is not just one less wafer for phones, it is effectively three fewer wafers worth of conventional memory.
• 2:31
  SPEAKER_00: That multiplier explains a lot about how fast this happened.
• 2:34
  SPEAKER_01: Right, and the shares have flipped dramatically.
• 2:37
  SPEAKER_01: Reporting from Tech Times estimates that AI data centers were consuming roughly 70% of all memory chips produced worldwide by the middle of this year, up from somewhere around 20 to 30% back in 2022.
• 2:49
  SPEAKER_00: That is a stunning shift in just four years.
• 2:52
  SPEAKER_00: And I assume the money behind it is enormous because nobody reorganizes an entire global chip industry without a very large check being waived at them.
• 3:01
  SPEAKER_01: The check is gigantic.
• 3:03
  SPEAKER_01: American hyperscalers, ThinkAlphabet, Amazon, Microsoft, Meta, and Oracle spent around$470 billion on capital expenditure last year.
• 3:12
  SPEAKER_01: This year it is about$870 billion, and next year it is projected to top$1.3 trillion.
• 3:18
  SPEAKER_00: When you put it next to a phone that sells for$90, the imbalance is almost comical.
• 3:23
  SPEAKER_00: A memory maker's choosing between a hyperscaler with a trillion dollars and a handset brand counting pennies on every unit.
• 3:29
  SPEAKER_01: And for that handset brand, memory is not a small line item.
• 3:33
  SPEAKER_01: Umdia says memory now accounts for more than 64% of the bill of materials and phones under$99.
• 3:40
  SPEAKER_01: So when memory prices jump, the entire phone becomes unprofitable.
• 3:44
  SPEAKER_00: 64% of the part's cost is just memory?
• 3:47
  SPEAKER_00: That reframes the whole thing for me.
• 3:49
  SPEAKER_00: Because you cannot simply trim the camera or the plastic and absorb a shock of that size.
• 3:54
  SPEAKER_01: No, you really cannot.
• 3:55
  SPEAKER_01: Carl Pay, the co-founder of Nothing, summed it up bluntly, saying brands now face a simple choice: raise prices by 30% or more or downgrade the specs.
• 4:04
  SPEAKER_01: And on the cheapest phones, there is almost nothing left to downgrade.
• 4:08
  SPEAKER_00: So what are manufacturers actually doing in practice?
• 4:11
  SPEAKER_00: I imagine the big Chinese brands, the ones that really built this low-end market, are the ones feeling the most heat right now.
• 4:17
  SPEAKER_01: They are.
• 4:18
  SPEAKER_01: Chinese manufacturers account for about 60% of global smartphone shipments.
• 4:23
  SPEAKER_01: And Ivan Lam at Counterpoint Research says many of them have drastically reduced their entry-level projects.
• 4:29
  SPEAKER_01: The most extreme number I found is that sub-100 shipments from Apo in Southeast Asia fell 96%.
• 4:36
  SPEAKER_00: 96% means they basically walked away from that tier in an entire region.
• 4:41
  SPEAKER_00: That is not a pricing adjustment anymore.
• 4:43
  SPEAKER_00: That is a strategic retreat from a whole class of customer.
• 4:46
  SPEAKER_01: And it is spilling into prices across the board.
• 4:48
  SPEAKER_01: Global smartphone prices rose about 15% this year, and newly launched models are around 25% more expensive than last year's launches.
• 4:57
  SPEAKER_01: Ivan Lamb put it memorably, saying what used to be below$150 may become below$250 or even$300.
• 5:05
  SPEAKER_00: Here's the part that jumps out at me.
• 5:07
  SPEAKER_00: I would guess those increases are not evenly spread, and that wealthy markets are being shielded while poorer ones take the hardest hit.
• 5:14
  SPEAKER_01: Your instinct is right.
• 5:16
  SPEAKER_01: Prices rose about 21% in India, 19% across Asia Pacific, 18% in the Middle East and Africa, and only about 5% in the United States.
• 5:26
  SPEAKER_01: The pain lands precisely where people have the least room to absorb it.
• 5:31
  SPEAKER_00: That is a really uncomfortable pattern.
• 5:33
  SPEAKER_00: The countries building the data centers barely notice the price change, while the countries that depend on cheap phones to get online at all are paying the steepest premium.
• 5:42
  SPEAKER_01: And it has shown up in the overall market, too.
• 5:45
  SPEAKER_01: The second quarter of this year was the worst second quarter for smartphone shipments since 2013.
• 5:50
  SPEAKER_01: Counterpoint measured an 11% decline.
• 5:53
  SPEAKER_01: IDC counted about 277 million units, down 6.7%, and Omdia saw a smaller drop of 4%.
• 6:02
  SPEAKER_00: I like that you gave all three numbers because the firms disagree on the size, but every single one of them agrees on the direction.
• 6:08
  SPEAKER_00: And that consistency is really the signal worth paying attention to.
• 6:12
  SPEAKER_01: Exactly.
• 6:12
  SPEAKER_01: And India is a great example of the speed.
• 6:15
  SPEAKER_01: Tech Times reported that the sub-100 segment in India collapsed by 59% year over year in the first quarter before the global numbers even caught up with it.
• 6:24
  SPEAKER_00: That timing is interesting because it suggests the budget segment acts like an early warning system.
• 6:30
  SPEAKER_00: The cheapest phones feel the squeeze first, long before anyone notices it in the premium flagships most analysts obsess over.
• 6:37
  SPEAKER_01: That is a sharp way to see it.
• 6:39
  SPEAKER_01: A flagship that costs$1,000 can hide a$40 memory increase inside its margin, but a$90 phone cannot hide anything, so it simply stops being made.
• 6:49
  SPEAKER_00: Let's go to Africa because I know that is where the budget phone is not just one option among many.
• 6:54
  SPEAKER_00: It is basically the whole market for a lot of people buying their first device.
• 6:58
  SPEAKER_01: That is exactly the right framing.
• 7:00
  SPEAKER_01: In Africa, 81% of smartphones shipped cost under$200.
• 7:05
  SPEAKER_01: Omdia found that overall shipments there fell 7% in the second quarter, but sub-100 shipments fell 34%, the sharpest decline of any price band.
• 7:15
  SPEAKER_00: And what does that mean for an actual household?
• 7:17
  SPEAKER_00: Put it in terms of income, because a percentage drop in shipments can feel very abstract to people listening from a comfortable office.
• 7:25
  SPEAKER_01: Here's the number that stopped me cold.
• 7:27
  SPEAKER_01: For the poorest fifth of the population, an entry-level device already costs a big share of monthly income, and in sub-Saharan Africa, it is roughly three-quarters of a month's earnings.
• 7:37
  SPEAKER_01: Raise that price by 20%, and the phone simply moves out of reach.
• 7:41
  SPEAKER_00: Which matters enormously because the phone is not a luxury in those markets.
• 7:45
  SPEAKER_00: It is the bank, the job board, the school, the health clinic, and increasingly the doorway to AI tools themselves.
• 7:52
  SPEAKER_01: That is the irony that makes this story so striking to me.
• 7:55
  SPEAKER_01: The GSMA projects that around 800 million more people will start using mobile internet by 2030, and many of them would reach AI assistance for the very first time through exactly these cheap phones.
• 8:07
  SPEAKER_00: So the AI buildout could end up pricing out the very people who might benefit the most from AI.
• 8:12
  SPEAKER_00: That is a genuinely painful loop, and I suspect most people in the industry have not thought about it at all.
• 8:16
  SPEAKER_01: Claire Sibthorpe from the GSMA Foundation said it clearly, warning that the spread of technology such as AI is creating greater digital divides.
• 8:25
  SPEAKER_01: She also noted that when phones stay available but become unaffordable, people make difficult sacrifices just to keep themselves online.
• 8:32
  SPEAKER_00: Okay, let me push back a bit because I do not want us to sound purely doom and gloom.
• 8:37
  SPEAKER_00: Memory markets are famously cyclical, so is there not a decent chance this sorts itself out once new factories come online?
• 8:44
  SPEAKER_01: That is a fair challenge, and the cycle argument has history on its side, but IDC's view is sobering.
• 8:50
  SPEAKER_01: Even after the shortage stabilizes around 2028, it expects the sub-100 segment to remain economically unviable.
• 8:58
  SPEAKER_01: Older memory types like LPDDR4X are being discontinued, not just deprioritized.
• 9:05
  SPEAKER_00: That is the detail that changes my mind.
• 9:08
  SPEAKER_00: If the cheap memory those phones relied on is literally being retired, then even a healthier market does not bring back the old$100 price point.
• 9:16
  SPEAKER_01: Right, the floor has moved, and the industry is adapting in interesting ways.
• 9:20
  SPEAKER_01: In Africa, Omdia says vendors are shifting from discounting to financing, partnering with companies like MCopa, Watu, and SafariCom, so people can pay off a phone in small installments instead of buying it outright.
• 9:33
  SPEAKER_00: Financing helps people get a device today, but it also means a lot of first-time buyers start their digital lives carrying debt.
• 9:40
  SPEAKER_00: I have mixed feelings about that trade, even if it is clearly better than having no phone.
• 9:44
  SPEAKER_01: I share that tension.
• 9:46
  SPEAKER_01: There is also a push to move buyers up toward the$120 to$150 band, where 5G becomes viable.
• 9:53
  SPEAKER_01: Transhin, the brand that dominates Africa, is holding its lead mostly through distribution reach and very disciplined inventory management.
• 10:01
  SPEAKER_00: So the winners in this new reality are the companies with logistics and lending muscle, not necessarily the ones with the cheapest bill of materials.
• 10:08
  SPEAKER_00: That is a very different competitive game from the one we had five years ago.
• 10:13
  SPEAKER_01: Precisely.
• 10:13
  SPEAKER_01: And it raises a question for the AI companies themselves.
• 10:17
  SPEAKER_01: If you are spending hundreds of billions on infrastructure to reach every person on Earth, you should probably care whether those people can still afford a device to reach you.
• 10:26
  SPEAKER_00: That is a great point.
• 10:27
  SPEAKER_00: And it suggests some real options.
• 10:29
  SPEAKER_00: Hyperscalers could fund device programs, or memory makers could reserve a slice of capacity for entry-level products, the way some industries protect essential supply during shortages.
• 10:39
  SPEAKER_01: I would add one more lever, which is software efficiency, phones that run well on less memory, lighter operating systems, smaller on-device models, all of that directly lowers the bill of materials and keeps more people connected despite expensive chips.
• 10:53
  SPEAKER_00: So for business leaders listening, I think the takeaway is that AI infrastructure is not an isolated story.
• 11:00
  SPEAKER_00: It touches consumer electronics, emerging market growth, and even the size of your future customer base in very concrete ways.
• 11:08
  SPEAKER_01: And for anyone planning products in India, Southeast Asia, or Africa, model your pricing assumptions carefully.
• 11:14
  SPEAKER_01: The device your customers use next year may cost far more than you expect, and that changes adoption curves for every digital service.
• 11:22
  SPEAKER_00: Laura, if you had to sum up today's story in one sentence for someone who missed everything we just said, what would you want them to remember most clearly from this conversation?
• 11:30
  SPEAKER_01: I would say that the memory feeding AI data centers is the same memory that used to make phones cheap, and right now the people with the least money are quietly paying for the AI boom with their access to the internet.
• 11:42
  SPEAKER_00: That is a powerful way to put it, and a reminder that every compute decision has ripple effects far beyond the data center.
• 11:48
  SPEAKER_00: Thank you, Laura.
• 11:49
  SPEAKER_00: This was a really eye opening conversation.
• 11:51
  SPEAKER_00: That's all for today's episode of the DX Today Podcast.
• 11:54
  SPEAKER_00: Thanks for listening, and we'll see you next time.