Episode 4

full
Published on:

11th May 2026

The Hidden Costs of the AI Boom on Consumer Electronics

The Hidden Costs of the AI Boom on Consumer Electronics

This week, we look at how the 2026 AI infrastructure boom is diverting silicon, memory production, and electricity away from consumer electronics, raising prices and worsening digital inequality.

We look at Bloomberg’s report that Alphabet, Amazon, Meta, and Microsoft budget about $650B in 2026 capex, dwarfing other industries, while the memory market’s three dominant suppliers (Samsung, SK Hynix, Micron) pivot toward high-bandwidth memory for AI.

OpenAI’s Stargate agreements are described as consuming up to 900,000 DRAM wafer starts per month (~40% of global output), and Micron exits consumer memory (ending Crucial). TrendForce projects steep 2026 price spikes for DRAM and NAND, driving IDC forecasts of falling smartphone and PC shipments, higher device prices, downgraded specs, and strained gaming GPU supply.

We also discuss the link between data centre's to rising electricity bills and US tariffs to further price increases, with refurbished devices as a limited stopgap and relief unlikely before 2027–2028.

Transcript
Speaker:

You are listening to Smarter

Articles, long form writing on

2

:

technology, governance, and the

human cost of the things we build.

3

:

This week's article is The Hidden Costs

Of The AI Boom On Consumer Electronics.

4

:

Somewhere in a fabrication facility

in Pyeongtaek, South Korea, a silicon

5

:

wafer that might have become the

memory in your next phone is being

6

:

sliced, stacked, and sold it into

something called high bandwidth memory.

7

:

It will never see the inside of a phone.

8

:

It will be bolted onto a server, GPU,

racked into a data center and fed

9

:

electricity until it helps generate a

response to someone's query about their

10

:

holiday plans or their quarterly report.

11

:

That wafer and millions like it

has been conscripted and you, the

12

:

person who just wanted a reasonably

priced laptop, are paying for it.

13

:

The numbers behind this transformation

are not easily absorbed.

14

:

In early 2026, Bloomberg reported

that four companies, Alphabet, Amazon,

15

:

Meta, and Microsoft have collectively

budgeted roughly $650 billion in

16

:

capital expenditure for the year.

17

:

Amazon alone.

18

:

$200 billion.

19

:

Alphabet, 185 billion.

20

:

Meta up to 135 billion.

21

:

Microsoft rounding out the

quartet at $105 billion.

22

:

Bloomberg's analysis of 21 other

major corporations spanning auto

23

:

making, aerospace and defense,

found their combined:

24

:

budgets totaling just $180 billion.

25

:

The AI infrastructure spend of

four Silicon Valley firms dwarfs

26

:

the capital plans of almost

every other industry on Earth.

27

:

Combined, this represents a 60% leap

from what those same four companies spent

28

:

in 2025 and 165% increase from 2024.

29

:

It is - in Bloomberg's own words - a

boom without a parallel this century.

30

:

That framing sounds celebratory.

31

:

It isn't, or at least it needn't

be because the same silicon, the

32

:

same fabrication lines, and the

same raw materials that power

33

:

everyday devices are being hoovered

up to feed these data centers.

34

:

The consequences are already visible

in electronic shops everywhere.

35

:

They are likely to get

worse before they improve.

36

:

The global memory chip market is an

oligopoly three manufacturers: Samsung

37

:

Electronics, SK Hynix, and Micron

Technology control virtually all of the

38

:

world's DRAM and NAND flash production.

39

:

When these three companies pivot

their manufacturing capacity in a

40

:

new direction, there is no fallback.

41

:

There is no alternative supplier

waiting quietly in Taiwan or Germany.

42

:

There is simply less memory

available for everything else.

43

:

That pivot is now well underway.

44

:

In October 2025, OpenAI signed agreements

with Samsung and SK Hynek to supply

45

:

memory chips for its Stargate Project.

46

:

A $500 billion AI infrastructure

program launched alongside SoftBank,

47

:

Oracle, and Abu Dhabi's MGX.

48

:

The scale was breathtaking.

49

:

Up to 900,000 DRAM wafer starts per month.

50

:

A volume that TrendForce estimated

could account for approximately

51

:

40% of total global DRAM output.

52

:

The deal was struck at the highest levels

of government and industry in a meeting

53

:

attended by Open AI's Chief Executive,

Samsung's Executive Chairman, SKS

54

:

Chairman, and South Korea's own President.

55

:

These were not procurement meetings.

56

:

These were geopolitical events.

57

:

In December, 2025, micron

clarified the picture even further.

58

:

The company announced it would

completely exit the consumer

59

:

memory market, discontinuing

its 29-year-old Crucial brand.

60

:

Micron's Chief Business Officer stated

plainly that the AI driven growth in

61

:

data centers had led to a surge in

demand and that the company had made the

62

:

decision to exit the consumer business

in order to support its larger strategic

63

:

customers in faster growing segments.

64

:

One of the three companies that

manufactures virtually all of the

65

:

world's memory had simply decided

that selling to ordinary people

66

:

was no longer worth the bother.

67

:

Micron reported record revenue of

billion in fiscal:

68

:

AI and data center applications

accounting for 56% of the total.

69

:

The economics were unambiguous.

70

:

Why sell thin margin RAM sticks to

consumers, when AI customers will pay a

71

:

premium for every wafer you can produce?

72

:

SK Hynek confirmed that its entire

dram NAND and HBM production through

73

:

2026 had been sold out, much of it

committed to NVIDIA for AI accelerators.

74

:

Samsung expanded its advanced

DRAM capacity, specifically

75

:

targeting HBM4 production.

76

:

The pattern is unmistakable.

77

:

Every major memory manufacturer

is reallocating capacity away from

78

:

consumer products and towards the

insatiable demands of AI infrastructure.

79

:

The physics of the problem makes

the trade off even starker.

80

:

HBM production for AI accelerators

consumes approximately three

81

:

times the wafer capacity of

standard DRAM per gigabyte.

82

:

This is a zero sum game.

83

:

Every wafer allocated to a stack for

an Nvidia GPU is a wafer denied to

84

:

the module in a mid-range smartphone

or the SSD in a consumer laptop.

85

:

Samsung and SK Hynix have both announced

plans to wind down DDR4 production.

86

:

China's ChangXin has reportedly ended

most of its own DDR4 production.

87

:

The supplier's tightening at

both the cutting edge and the

88

:

budget end simultaneously.

89

:

The impact on memory prices has

been nothing short of historic.

90

:

TrendForce projected that conventional

DRAM contract prices would surge

91

:

by 90 to 95% quarter on quarter

in the first quarter of:

92

:

NAND flash prices were expected to

rise by 55 to 60% in the same period.

93

:

PC DRAM specifically was projected

to increase by over a hundred

94

:

percent in a single quarter.

95

:

A new record for the steepest

quarterly surge ever recorded in

96

:

the memory industry's history.

97

:

DRAM's spot prices increased 172% year

ar as of the third quarter of:

98

:

Retail prices for 32 gigabyte DDR5

modules jumped somewhere between

99

:

163% and 619% in global markets

since September of that year.

100

:

One Terabit TLC NAN devices climbed

from roughly $4 and 80 cents in

101

:

July, 2025 to around $10 and 70

cents by late in the year, more

102

:

than doubling in barely six months.

103

:

The root cause is

structural, not cyclical.

104

:

Unlike previous memory price spikes driven

by earthquakes or temporary mismatches

105

:

between supply and demand, this shortage

reflects a deliberate and potentially

106

:

permanent strategic reallocation of

the world's silicon wafer capacity.

107

:

Phison's chief executive told industry

publications that every NAND manufacturer

108

:

had confirmed 2026 was sold out.

109

:

Silicon Motions Chief Executive

offered an even more sobering summary.

110

:

What the industry faces he

said has never happened before.

111

:

HDD, DRAM, HBM and NAND all in

severe shortage simultaneously.

112

:

NAND vendors remain cautious about adding

fabrication capacity after several years

113

:

of weak profitability with new production

lines delayed until at least:

114

:

Relief is not imminent.

115

:

Relief is not soon.

116

:

The downstream effects on consumer devices

are already visible and they're grim.

117

:

IDC warned in February, 2026 that the

global smartphone market is poised

118

:

to suffer its biggest decline ever

with shipments expected to drop 12.9%

119

:

to 1.12

120

:

billion units the lowest

level in more than a decade.

121

:

The average selling price of

smartphones is projected to

122

:

surge 14% to a record $523.

123

:

For budget conscious buyers,

the picture is worse still.

124

:

Counterpoint Research found that the bill

of materials cost for low-end smartphones

125

:

price below $200 has increased 20 to

30% since the beginning of the year.

126

:

IDC warned that the sub $100

smartphone segment representing 171

127

:

million devices annually will become

permanently uneconomical even after

128

:

memory prices stabilized by mid 2027.

129

:

Those devices are not a niche product.

130

:

They are the on-ramp to the

digital economy for hundreds of

131

:

millions of people across Africa,

south Asia, and Southeast Asia.

132

:

When that ramp is pulled away, the

promise that AI will benefit all of

133

:

humanity begins to ring rather hollow.

134

:

Some manufacturers are responding

with a quiet downgrade strategy that

135

:

consumers may not immediately notice.

136

:

A 2026 mid-range smartphone

might ship with six gigabytes

137

:

of ramp, where it's 2025.

138

:

Predecessor offered eight gigabytes.

139

:

At the low end base models are

likely to return to four gigabytes.

140

:

A specification most consumers associate

with phones from several years ago.

141

:

The model name stays the same,

the marketing stays the same.

142

:

You are getting less for more.

143

:

The irony is sharp.

144

:

The technology industry has spent the past

two years marketing AI smartphones devices

145

:

with enhanced on-device AI capabilities

that typically require more ram, not less.

146

:

Now the very infrastructure being

built to power those AI models is

147

:

cannibalising the memory supply.

148

:

Those phones need to run them.

149

:

The personal computer market faces a

similarly painful reckoning Memory now

150

:

accounts for around 20% of the hardware

costs of a laptop up from between 10

151

:

and 18% in the first half of 2025.

152

:

Lenovo, Dell, hp, ASA and asis have

all warned of price hikes of 15 to

153

:

20% as an industry-wide response.

154

:

IDC warned that the PC market

% in:

155

:

Gartner projected that rising

memory prices will make low margin

156

:

entry level laptops under $500,

financially unviable within two years.

157

:

The timing could hardly be worse.

158

:

The memory shortage has collided

with Microsoft's Windows 10 end

159

:

of Lifecycle, which was supposed

to drive a major refresh wave as

160

:

consumers and businesses upgraded.

161

:

Instead, the very components needed

to build those new machines are being

162

:

siphoned off to fill AI server racks.

163

:

The planned A IPC marketing push,

which was meant to entice consumers

164

:

with on-device AI capabilities.

165

:

Requiring more RAM now faces the bitter

irony that AI's own infrastructure demands

166

:

have made that extra memory unaffordable.

167

:

The feature and the shortage

share the same cause.

168

:

PC gaming enthusiasts a community

already accustomed to volatility

169

:

in component pricing are facing

yet another punishing cycle.

170

:

MSI President described 2026

as the most difficult year

171

:

since the company was founded.

172

:

NVIDIA's GForce, RTX 5,080 has

experienced price increases of up to 35%.

173

:

The flagship RTX 5,090 has

seen a 79% price increase.

174

:

A MD has told supply partners

it will raise graphics card

175

:

prices by at least 10%.

176

:

Reports suggest major graphics

card makers may trim production of

177

:

consumer lines by 30 to 40% in 2026.

178

:

Nvidia reportedly has no plans

to release any new GForce gaming

179

:

graphics cards until 2027.

180

:

The underlying cause is the same memory

shortage affecting phones and laptops.

181

:

Every wafer diverted to an AI

accelerator is a wafer that

182

:

never reached the consumer shelf.

183

:

The memory chip shortage is only one

vector through which AI infrastructure

184

:

costs are reaching ordinary consumers.

185

:

There is another less visible but

equally consequential channel.

186

:

Electricity data centers accounted for

around one and a half percent of the

187

:

world's electricity consumption in 2024.

188

:

Gartner estimates that worldwide data

center electricity consumption will rise

189

:

from 448 TERAWATT hours in 2025 to 980

terawatt hours by:

190

:

servers, electricity usage set to rise,

nearly fivefold in the same period.

191

:

A January, 2026 report predicted that

American Data Centers total combined

192

:

energy demand will nearly double between

:

193

:

needs of Spain in just three years.

194

:

This is not an abstract infrastructure

concern in the PJM electricity market.

195

:

Stretching from Illinois

to North Carolina.

196

:

Data centers accounted

for an estimated $9.3

197

:

billion price increase in the

:

198

:

As a result, the average residential bill

is expected to rise by $18 a month in

199

:

Western Maryland and $16 a month in Ohio.

200

:

A Carnegie Mellon University study

estimates that data centers and

201

:

cryptocurrency mining could lead

to an 8% increase in the average

202

:

American electricity bill by 2030.

203

:

Potentially exceeding 25% in

the highest demand markets of

204

:

Northern Virginia in Ireland.

205

:

Around 21% of all electricity is

already consumed by data centers.

206

:

A survey found that 78% of

Americans are somewhat or very

207

:

concerned that new data centers

will make their energy bills go up.

208

:

Those concerns are well-founded and then.

209

:

Compounding the memory shortage

and the rising energy bills.

210

:

There are tariffs in the United States.

211

:

Sweeping trade policy changes have

imposed significant duties on key

212

:

technology manufacturing partners,

including a 30% tariff on Chinese goods,

213

:

and a 20% duty on Vietnamese imports.

214

:

Analysis by the Consumer Technology

Association found that these tariffs could

215

:

result in smartphone prices increasing by

31%, laptop and tablet prices rising 34%,

216

:

and gaming console prices jumping 69%.

217

:

The CTA estimated that for every dollar

in gains to domestic producers, consumers

218

:

may lose up to $16 in spending power.

219

:

Each pressure alone would be significant.

220

:

Together they represent a fundamental

repricing of everyday technology felt most

221

:

acutely by those who can least afford it.

222

:

The World Bank's 2025 Digital Progress

and Trends report noted that high

223

:

income countries host 77% of global

co-location data center capacity

224

:

while lower middle income countries

hold just 5% and low income countries

225

:

less than a 10th of a percent.

226

:

Africa accounts for less than 1% of

global data center capacity, despite being

227

:

home to 18% of the world's population.

228

:

High income countries account for 87% of

notable AI models, 86% of AI startups,

229

:

and 91% of venture capital funding.

230

:

Despite representing just 17% of the

global population, approximately 2.2

231

:

billion people remain offline.

232

:

Mostly in low and middle income countries.

233

:

In 2024, a basic broadband plan

consumed 29% of monthly income in

234

:

low income countries compared with

less than 3% in high income ones.

235

:

When the price of the device is needed

to get online rises, 20 to 30% because

236

:

memory chips are being diverted to

data centers in Virginia and Oregon.

237

:

The impact on digital

inclusion is severe, immediate.

238

:

Largely unremarked upon in

the places where those data

239

:

centers are being celebrated.

240

:

One unexpected beneficiary of all

this is the refurbished electronics

241

:

market, which is growing quickly

as consumers seek alternatives to

242

:

increasingly expensive new devices

valued at around $130 billion globally

243

:

in 2025, growing at over 11% annually.

244

:

Nearly one in seven smartphones sold

rance in the first quarter of:

245

:

Were refurbished in Britain.

246

:

Close to 10% of all smartphones

sold in the same period.

247

:

Were secondhand.

248

:

But the refurbished market is

ultimately a stop gap, not a solution.

249

:

It depends on a steady

flow of traded in devices.

250

:

If new device sales decline as

sharply as IDC projects, the

251

:

supply available for refurbishment

will eventually thin as well.

252

:

Secondhand markets cannot

absorb the shortfall of a

253

:

market in structural retreat.

254

:

Relief from the memory shortage is not

expected until:

255

:

new fabrication facilities from Samsung

and SK Hynek reach volume production.

256

:

Samsung's new Pyeongtaek facility

is expected operational by:

257

:

Micron is building factories in Idaho

ill start producing memory in:

258

:

and 2028, but even when new capacity

arrives, there is no guarantee it will

259

:

be allocated to consumer products.

260

:

If AI demand continues at its current

trajectory, and if the economic

261

:

incentives continue to favor high

margin enterprise customers over

262

:

consumer markets, the structural

reallocation may simply persist.

263

:

TrendForce does not expect DRAM prices

to decline at any point in:

264

:

senior research director has been blunt.

265

:

This is not a temporary decline.

266

:

This is a structural reset

of the entire market.

267

:

There is a deep irony at

the heart of this story.

268

:

The technology industry has spent

three years telling us that artificial

269

:

intelligence will transform our

lives, democratize access to

270

:

information, and solve problems that

have long eluded human ingenuity.

271

:

Some of that may prove true, but

right now, in:

272

:

and measurable impact of the AI

boom on ordinary people is this.

273

:

Your phone costs more, your laptop costs

more, your graphics card costs more.

274

:

Your electricity bill is rising, and the

cheapest devices that connect hundreds of

275

:

millions of people in the developing world

to the internet are becoming economically

276

:

unviable because of AI's own appetite

for the resources needed to build them.

277

:

The $650 billion being poured

into data centers this year

278

:

is not coming from nowhere.

279

:

It is being extracted indirectly,

but inexorably from the

280

:

consumer technology ecosystem.

281

:

The fabrication lines that once produced

your memory chips now produce ai memory.

282

:

The electricity that once powered your

neighborhood now powers server farms.

283

:

The manufacturing capacity that once

kept entry-level devices affordable is

284

:

committed for years ahead to contracts

with hyperscale cloud providers.

285

:

The memory industry's oligopolistic

structure means that decisions made in

286

:

a handful of boardrooms in Seoul, Boise,

and itch on ripple outward to affect

287

:

the price of every device on the planet.

288

:

When a single project can sign agreements,

consuming 40% of global DRAM output.

289

:

When a single company can exit the

consumer memory market entirely, because

290

:

AI customers are simply more profitable,

and when entry level devices for billions

291

:

of people become permanently uneconomical,

the market is sending a clear signal.

292

:

The benefits of the AI boom accrue to

the companies building it and eventually

293

:

to the users of their products.

294

:

The costs are being socialized across

the entire consumer technology market

295

:

in higher prices, reduce specifications.

296

:

Rising energy bills and a widening

digital divide, the people least likely

297

:

to benefit from advanced AI models are

the same people most affected by the

298

:

rising price of the devices they need to

participate in the digital economy at all.

299

:

That is not an accident of the market.

300

:

It is the market working

exactly as intended.

301

:

You've been listening to Smarter Articles.

302

:

The article you just heard was first

published at smarterarticles.co.uk,

303

:

where you'll find our full

archive, a new article every day.

304

:

Thanks for listening.

305

:

Subscribe wherever you get your podcasts.

306

:

Share with someone who thinks carefully

and we'll meet here again next week.

All Episodes Previous Episode

Listen for free

Show artwork for SmarterArticles

About the Podcast

SmarterArticles
Keeping the Human in the Loop
A weekly audio edition of the long-running independent journal. Each bulletin brings carefully argued pieces on artificial intelligence, decentralised cognition, posthuman ethics, and the quiet politics of the technologies reshaping daily life.

AI voice narration from ElevenLabs Studio is used in the production of this Podcast.

About your host

Profile picture for Tim Green

Tim Green

UK-based Systems Theorist & Independent Technology Writer

Tim explores the intersections of artificial intelligence, decentralised cognition, and posthuman ethics. His work, published at smarterarticles.co.uk, challenges dominant narratives of technological progress while proposing interdisciplinary frameworks for collective intelligence and digital stewardship.

His writing has been featured on Ground News and shared by independent researchers across both academic and technological communities.

ORCID: 0009-0002-0156-9795