Average DRAM price in USD over last 18 months

📝 Discussion Summary (Click to expand)

The three most prevalent themes in the Hacker News discussion are:

The Impact of AI Demand on Component Supply Chain and Pricing: There is significant discussion and concern that massive purchases by AI companies (specifically OpenAI hoarding DRAM wafers) are directly causing the current price surges and supply constraints for memory available to the general consumer and PC market.
- Supporting Quote: "More specifically, it's OpenAI quietly contracting away 40% of global DRAM capacity, only to build wafers of memory and shove it in a warehouse." ("delfinom")
Frustration Over Inflation and Skewed Economic Metrics: Several users express anger that while essential goods (like housing, insurance, and food) have seen dramatic price increases, official inflation statistics fail to capture the felt impact on typical consumers. The high price of technology components like RAM is seen as part of this broader economic squeeze.
- Supporting Quote: "Supposedly $1 in 2000 is worth $1.88 in 2025. So 88% inflation over the 25 years. Meanwhile the median home price has increased by 150%." ("tstrimple")
Skepticism Regarding the Future Value and Liquidation of AI Hardware: There is debate over what would happen to expensive, specialized AI hardware (like H100s or massive DRAM stockpiles) if the "AI bubble" were to burst, with opinions split between rapid depreciation and eventual secondary market repurposing.
- Supporting Quote: "Unless it drags on for years so that the parts are old by the time everything gets liquidated." ("gblargg")

🚀 Project Ideas

A cloud-based or local service to stress-test and certify used DRAM modules for specific tasks (e.g., standard workstation use, NAS, or even AI inference) despite minor reported errors.
This directly addresses the need for reliable used hardware, especially addressing the fact that "memory sticks just sitting around" might have intermittent failures (like a bad memory channel mentioned by willis936 or modules failing Memtest86 some of the time mentioned by thescriptkiddie).
Core Value Proposition: Providing a quantifiable, time-stamped reliability score for second-hand components, moving them from "worthless" to certifiably usable for certain use cases.

Key	Value
Target Audience	E-waste recyclers, second-hand hardware resellers (eBay/AliExpress sellers), tech hobbyists building budget systems seeking reliability guarantees.
Core Feature	Automated execution of memory-intensive diagnostics (e.g., modified Memtest variants, custom read/write patterns) over extended periods, followed by a tiered certification report (e.g., "Tested Stable for 48h Workstation Use").
Tech Stack	Rust/Go for high-speed testing harness, containerization (Docker/Podman) for sandboxing tests, Web platform (Next.js/FastAPI) for submission and report delivery.
Difficulty	Medium (Requires deep understanding of memory controllers, kernel parameters like `memmap` mentioned by `sznio`, and hardware setup automation).
Monetization	Hobby

Users expressed interest in using "semi-reliable RAM for training" (krackers). This service allows sellers to test modules down to a known error profile, enabling specialized sales to those willing to patch kernels or run tests (sznio, viraptor).
The core value is establishing trust: "What I have now is certified good enough for a specific load." This helps counter the general dismissal of used server/RAM parts as being in a "bad state of wear" (magarnicle).

A specialized market data platform that tracks the supply, projected liquidation schedule, and real-time pricing of high-value, surplus AI hardware (H100s, H200s, enterprise DRAM, specialized network cards).
Solves the need for individuals who want to acquire potentially deflationary assets ("I would snag up every possible H100... if the AI bubble burst") and understand the timeline for asset depreciation.
Core Value Proposition: Providing actionable investment intelligence on the volatile secondary market for compute hardware, bridging the gap between current speculation and future availability.

Key	Value
Target Audience	Speculators (`edm0nd`), IT asset disposal (ITAD) firms, large system integrators, hobbyists planning long-term infrastructure builds (e.g., budget cracking rigs).
Core Feature	Aggregation of known/rumored large-scale asset turnover events, dynamic modeling predicting when server/DRAM saturation will cause steep price drops, and alerts when hardware hits specified low-price thresholds.
Tech Stack	Python (Scrapy for scraping, Pandas for data modeling), PostgreSQL for time-series data, Machine Learning models for predictive timing (leveraging comments on asset aging like `Havoc` and `gblargg`).
Difficulty	High (Requires deep supply chain knowledge, interpreting corporate statements, and handling highly anecdotal data sources like specialized industry blogs and regulatory filings).
Monetization	Hobby

Directly addresses the desire to wait for the "AI bubble burst" and capitalize on low prices for potentially great rigs (e.g., "make cheap and fast password hash cracking rigs").
It navigates the uncertainty around when hardware will actually be liquidated versus when it’s just announced as purchased/in storage ("piling it in a warehouse" - deltoidmaximus), helping users avoid buying just before a cliff.

A tool that allows users to input custom spending weights (e.g., percentage spent on rent, education, insulin, vs. electronics like TVs/DRAM) to generate a personalized inflation rate, moving beyond the flawed "hypothetical average person" CPI basket.
Solves the frustration that official inflation metrics do not reflect felt reality for specific economic segments ("The inflation basket only represent a hypothetical average person" - energy123).
Core Value Proposition: Replacing generalized economic statistics with verifiable, personalized metrics that accurately reflect the cost of living for different lifestyles, allowing for better policy commentary and personal budgeting.

Key	Value
Target Audience	Economically engaged individuals, advocates, policy wonks, and users frustrated with standard cost-of-living adjustments (`tstrimple`, `energy123`).
Core Feature	User inputs customizable weights across major cost categories, pulling aggregated component price data (including housing/healthcare/eggs/Budweiser) to calculate a personalized inflation % and Gini coefficient shift measurement (`kevindavis`).
Tech Stack	Vue.js/Svelte for interactive front-end, Node.js/AWS Lambda for backend calculations, potentially linking to open government datasets and curated hardware price histories (like DDR4 prices).
Difficulty	Medium (Data aggregation across disparate sources like housing data and fluctuating DRAM spot prices is challenging, but the core calculation is straightforward).
Monetization	Hobby

This project directly appeals to the deep desire for better economic transparency seen in the thread: "It's more useful to construct multiple separate inflation measures that represent different types of people."
It moves the discussion past simple anecdotes by allowing users to quantify their individual experience, potentially showing negative inflation for asset owners and positive inflation for renters, as suggested by energy123.