--- title: "AWS GPU prices jump 20%, SimpleHelp RCE targets AI coding tokens" url: https://daily.dev/posts/aws-gpu-prices-jump-20-simplehelp-rce-targets-ai-coding-tokens-ysfcq17af source_url: https://daily.dev/posts/aws-gpu-prices-jump-20-simplehelp-rce-targets-ai-coding-tokens-ysfcq17af type: freeform source: "Backend Digest" published: 2026-06-30T04:18:40.652Z updated: 2026-06-30T04:18:57.359Z reading_time: 4 upvotes: 0 comments: 0 language: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # AWS GPU prices jump 20%, SimpleHelp RCE targets AI coding tokens **[Backend Digest](https://daily.dev/sources/backend_digest)** · 4 min read · 0 upvotes · 0 comments ## Summary A curated developer news digest covering five major stories: AWS raised EC2 Capacity Block prices for Nvidia GPUs ~20% due to HBM shortages; a critical SimpleHelp RMM authentication bypass (CVE-2026-48558) is being actively exploited to steal AI coding assistant tokens and cloud credentials; Uber is reportedly dropping PagerDuty after 12+ years; Valkey 9.1 lands on Amazon ElastiCache with 2.1M req/s throughput and new multi-tenant ACLs. Notable items include Securonix saving $135K/year migrating Redis clusters to Valkey, Deno 2.9 halving cold-start times, GitHub Advisory Database falling behind on vulnerability review volume, MCP-based symbol graphs boosting AI coding agent recall to 100%, and Argo CD OOM issues at scale compared to Sveltos. ## Content **TLDR:** AWS raised EC2 Capacity Block prices for Nvidia GPUs by roughly 20% starting July, the second hike in six months, driven by a high-bandwidth memory shortage. A critical authentication bypass in SimpleHelp RMM (CVE-2026-48558) is being actively exploited to steal tokens from AI coding assistants, cloud credentials, and package registry keys. Uber is reportedly dropping PagerDuty after 12+ years, a signal of broader industry migration away from the platform. Valkey 9.1 landed on Amazon ElastiCache with a redesigned I/O threading model hitting 2.1M requests/second and new multi-tenant ACLs. --- ## CVE-2026-48558 in SimpleHelp RMM actively exploited to steal AI coding assistant tokens Attackers are exploiting a critical authentication bypass in SimpleHelp RMM to deploy Djinn Stealer, which targets an unusually broad blast radius: tokens from Claude, Gemini, Codex, Cline, and OpenCode; Git/SSH credentials; Docker and Helm configs; npm, PyPI, Cargo, and Maven registry credentials; and AWS, Azure, and GCP keys. Because AI coding tools connect to repos, databases, and cloud accounts via MCP, a single compromised machine can expose the full development supply chain. RMM abuse jumped 277% year-over-year according to Huntress's 2026 Cyber Threat Report. If SimpleHelp is anywhere in your toolchain, patch or isolate it now. [Read more](https://daily.dev/posts/akTn6O3jr) ## AWS EC2 Capacity Block prices up 20% as HBM shortage tightens GPU supply AWS raised reserved Nvidia GPU block prices by roughly 20% starting July — the second increase in six months after a 15% hike in January. The root cause is a high-bandwidth memory shortage constraining GPU production globally, giving hyperscalers pricing power since customers have few alternatives. Trainium and on-demand purchasing options are unaffected for now. AI teams running large training or fine-tuning workloads on reserved capacity should model the cost trajectory; the same memory crunch is hitting Apple hardware and Xbox, while benefiting Micron and SK Hynix. [Read more](https://daily.dev/posts/JdminsUs0) ## Uber drops PagerDuty after 12+ years as incident management migration accelerates Uber is reportedly ending its relationship with PagerDuty, with multiple sources citing the same pattern: PagerDuty missed the Slack shift and then the AI shift after going public, and many tech companies are now actively migrating off. This isn't a single customer churn story — it's a signal about where incident management tooling is heading. If you're mid-contract with PagerDuty or evaluating alternatives, the market is clearly in motion. [Read more](https://daily.dev/feed-by-ids?id=Lv2cqv3Vg&id=mugO2Igkp&id=wnUPgBsSw) ## Valkey 9.1 on ElastiCache: 2.1M req/s, 20% memory reduction, new multi-tenant ACLs Amazon ElastiCache now supports Valkey 9.1, which ships a redesigned I/O threading model reaching 2.1M requests/second, up to 20% memory reduction for small strings, and up to 10% savings for sorted sets. New database-level ACLs enable stronger isolation for multi-tenant cluster deployments — a meaningful addition if you're sharing clusters across teams or services. Three new commands are worth knowing: HGETDEL for atomic hash field retrieval and deletion, MSETEX for setting multiple keys with a shared TTL, and CLUSTERSCAN for unified key scanning across a cluster. Available at no additional cost in supported AWS regions. [Read more](https://daily.dev/posts/pvEmzNhnh) --- ## Also notable - **Securonix migrated 250+ ElastiCache clusters from Redis to Valkey in two weeks, saving $135K/year:** Securonix moved 250+ clusters with zero downtime and no application code changes, achieving 20% cost reduction (~$135K annually) and 30%+ query performance improvement via Valkey's multi-threaded architecture. [Read more](https://daily.dev/posts/eN9rTuQ1K) - **Deno 2.9 cuts cold-start time from 34ms to 17ms and adds experimental desktop app support:** Deno 2.9 halves cold-start latency, reduces peak RSS by 2.2–3.1x under load, improves HTTP/1.1 throughput, and ships an experimental `deno desktop` command for building native apps as single distributable binaries. [Read more](https://daily.dev/posts/NLlbqv6xP) - **GitHub Advisory Database published 1,560 reviewed advisories in May 2026, still falling behind 3,000+/week incoming volume:** Private vulnerability reports surged from ~550/week in January to 3,000+/week in May, with GitHub CNA CVE requests hitting nearly 4,000 in May alone — review times have extended to multiple weeks for a meaningful share of submissions. [Read more](https://daily.dev/posts/qIkJEIvQp) - **AI coding agent recall on a large Rails codebase went from 0.09 to 1.00 with a symbol graph tool via MCP:** Without a structural code graph, agents found 2 of 11 scattered dependents in Chatwoot; with Sense (a codebase index served via MCP), all five models tested found all 11, ran 21% faster, and used 18% fewer tokens. [Read more](https://daily.dev/posts/XTMp02kvY) - **Argo CD hits OOM around 15k–20k cached objects per hub; Sveltos handled the same load at 2GB vs 21GB RAM:** A 31-iteration test across 15,000+ vClusters found Argo CD's application controller memory usage scales non-linearly with object count, not cluster count — at very large scale, Sveltos deployed 1,000 apps across 250 vClusters in 35 minutes with sharding. [Read more](https://daily.dev/posts/ABAZXoDax) ## Similar posts on daily.dev - [AWS GPU prices jump 20% as memory crunch bites](https://daily.dev/posts/aws-gpu-prices-jump-20-as-memory-crunch-bites-jdminsus0) · The Next Web · 0 upvotes · 0 comments - [AWS Hikes EC2 Capacity Block Rates by 15% in Uniform ML Pricing Adjustment](https://daily.dev/posts/aws-hikes-ec2-capacity-block-rates-by-15-in-uniform-ml-pricing-adjustment-lejupz6of) · InfoQ · 0 upvotes · 0 comments --- [View this post on daily.dev](https://daily.dev/posts/aws-gpu-prices-jump-20-simplehelp-rce-targets-ai-coding-tokens-ysfcq17af)