AI industry
Cheap APIs
↓
User growth
↓
Monetization
↓
Ecosystem
RESEARCH BRIEF / RB-01 · V1.0
RB-01 · The End of
Cheap Growth
What DeepSeek and China's Auto Industry Reveal
About the Next Industrial Era
About the Next Industrial Era
Cheap AI and cheap cars were once growth engines.
Now both are hitting the same inflection point — value, not price, decides what happens next.
The Question
What happens to an industry when its core input suddenly gets cheap?
Executive Summary
Two industries, one inflection point.
In August 2026, DeepSeek introduced peak/off-peak pricing; the same month, China’s auto retail fell -17% YoY. One is software, the other is manufacturing — but both point to the same turning point.
01
AI's inflection
DeepSeek’s peak/off-peak pricing is not a simple price hike — it marks AI’s move from user acquisition to sustainable infrastructure.
02
Spending is reallocating
China’s auto retail fell -13.2% YoY in Jan–Jul 2026 (-17% in July), while telecom equipment rose +20.4% — consumers haven’t stopped; they’ve changed where money goes.
03
Structural pressure
Overcapacity, price wars, and the redefinition of the car are eroding the old “scale = advantage” logic.
Step 1
Cheap user acquisition
Lower barriers, rapid adoption
Step 2
Unit-economics pressure
Compute and capacity costs surface
Step 3
Shift to profit & efficiency
Firms optimize cost structure
Step 4
Competition variables change
From price to technology, ecosystem
Step 5
Value restructuring
The less replaceable wins
01 — Peak/Off-Peak Pricing
AI is starting to price against the real cost curve of compute
DeepSeek made high-performance models remarkably cheap. This time, it rewrote the pricing model itself.
In August 2026, DeepSeek introduced peak/off-peak pricing for its V4 models: weekday peak hours (Beijing time 09:00–12:00 and 14:00–18:00) saw sharp increases, off-peak hours cost half the peak rate, and weekends are billed entirely off-peak. Prices went both up and down — not a one-way increase. This is a shift in the pricing mechanism: AI services are beginning to be priced against the real cost curve of computing.
Because AI has never been free. Behind every model are compute, GPUs, data centers, electricity, and engineering talent; the cheaper AI is for users, the more efficiency matters for the firms building it. Prices tracking the cost structure means the industry is moving from “how do we get more people to use AI” to “how do we build sustainable AI infrastructure.”
FIG. 01 — DeepSeek V4 Peak/Off-Peak Pricing: Up and Down
Unit: CNY per million tokens
| Model | Meter | Before | Peak | Increase | Off-peak · weekend |
|---|---|---|---|---|---|
| V4-Pro | Input · cache hit | 0.025 | 0.30 | ↑ 1100% | 0.15 |
| Input · cache miss | 3 | 9 | ↑ 200% | 4.5 | |
| Output | 6 | 27 | ↑ 350% | 13.5 | |
| V4-Flash | Input · cache hit | 0.02 | 0.10 | ↑ 400% | 0.05 |
| Input · cache miss | 1 | 3 | ↑ 200% | 1.5 | |
| Output | 2 | 9 | ↑ 350% | 4.5 |
Note: “Before” is the flat price prior to the August 17 adjustment. Peak hours are weekdays 09:00–12:00 and 14:00–18:00 Beijing time; off-peak is half the peak rate; from August 23, 2026, weekends are billed entirely off-peak.
So What
Peak/off-peak pricing is not “AI raising prices” — it is a signal that AI is entering commercialization, moving from user acquisition to sustainable infrastructure.
02 — Consumption Rebalancing
Consumers haven't stopped spending — they've changed where it goes.
Per China’s National Bureau of Statistics, total retail sales of consumer goods rose +1.2% YoY in Jan–Jul 2026, while automobile retail fell -13.2% — among the weakest major categories.
FIG. 02 — Auto Decline vs. Consumption Rebalancing
Source: NBS · July 2026
Auto retail
-13.2%
Jan–Jul 2026 YoY
Auto retail
-17.0%
July 2026 YoY
Telecom equipment
+20.4%
July 2026 YoY
Total retail sales
+1.2%
Jan–Jul 2026 YoY
Autos are among the weakest categories, but consumers haven’t stopped spending — they’ve changed where it goes: telecom equipment rose +20.4% in July, with food, cosmetics, and others still growing, while cars, building materials, jewelry, and furniture contracted. This is not vanishing consumption but a rebalancing — big-ticket purchases are deferred, while small, high-certainty spending continues.
FIG. 03 — Retail Sales Growth by Category, July 2026, YoY
Unit: %
Source: National Bureau of Statistics of China, “Total Retail Sales of Consumer Goods in July 2026,” released August 18, 2026. YoY growth, not adjusted for price factors; selected categories from retail enterprises (units) above the designated size.
So What
Reading auto’s decline as a “consumption winter” misses the real signal: consumption isn’t disappearing — it’s re-pricing priorities.
03 — The Structural Problem in Auto
The pressure is structural, not cyclical.
Attributing the slowdown to EV competition or the macro environment is incomplete. The deeper cause is three structural shifts.
I
Overcapacity
The “expand → lower costs → sell more” flywheel of the growth era becomes pressure once market growth slows: competition shifts from “who expands fastest” to “who operates most efficiently.”
II
The price-war backlash
Low prices pull sales forward in the short run but squeeze margins and weaken long-term R&D. Durable competitiveness comes from innovation, not discounts.
III
The car is being redefined
Competition is moving from engines, transmissions, and mechanical manufacturing to batteries, software, and autonomous driving — from mechanical product to intelligent mobile terminal.
So What
Winners won’t necessarily be the highest-volume producers, but companies with better technology, efficiency, and financial discipline.
04 — One Industry Pattern
AI and autos face the same question.
Why put AI and cars together? Because both face the same fundamental question: can low prices create sustainable value?
Early in an industry, low prices drive adoption, attract users, and accelerate growth; but every industry eventually reaches the same inflection point — price is no longer sufficient, and firms must compete on technology, efficiency, ecosystem, brand, and global capability.
FIG. 04 — Two Faces of the Same Question
AI vs. autos · key competitive variables
Auto industry
Cheap cars
↓
Market share
↓
Profitability
↓
Technology
So What
AI can’t compete forever on cheaper APIs, and autos can’t compete forever on discounts — both industries share the same pattern.
05 — Future Winners
Industries won't disappear — their rules are being rewritten.
Autos won’t disappear, and AI won’t slow down; what’s changing is the competitive landscape. In the coming years, many industries will re-sort.
FIG. 05 — The Winners: Companies and Individuals
Irreplaceability is the shared answer
Companies
Technology
↓
Efficiency
↓
Cash flow
↓
Global capability
Individuals
Domain knowledge
↓
Software skills
↓
AI capability
↓
Communication
What Could Change My Mind
If cheaper AI clearly grows the pie rather than just compressing margins — if adoption outpaces the price deflation — the “end of cheap growth” thesis would be wrong.
Conclusion
The end of cheap growth.
Value competition has begun.
The future belongs not to the cheapest, but to those who create genuinely irreplaceable value.
References
Evidence Sources
Core evidence draws on public 2026 data; growth rates are year-on-year and not adjusted for price factors.
01
National Bureau of Statistics of China — "Total Retail Sales of Consumer Goods in July 2026" Auto -13.2% / -17% · Telecom equipment +20.4% · Total retail +1.2%, released 2026-08-18 core
stats.gov.cn
02
DeepSeek API Docs — Models & Pricing V4 peak/off-peak pricing: price schedule and effective dates core
api-docs.deepseek.com
Open Questions
- Which industries will face the same "cheapness inflection" next?
- Does the value-not-price pattern from autos hold for AI — or differ?