Cloudflare status: hosting issues and outage reports

Problems in the last 24 hours

The graph below depicts the number of Cloudflare reports received over the last 24 hours by time of day. When the number of reports exceeds the baseline, represented by the red line, an outage is determined.

Most Reported Problems

The following are the most recent problems reported by Cloudflare users through our website.

• Domains (41%)
• Cloud Services (27%)
• Hosting (18%)
• Web Tools (9%)
• E-mail (5%)

Live Outage Map

The most recent Cloudflare outage reports came from the following cities:

City	Problem Type	Report Time
Angers	Cloud Services	2 days ago
London	Domains	4 days ago
Noida	Hosting	17 days ago
Jewar	E-mail	17 days ago
Braga	Web Tools	17 days ago
Noida	Cloud Services	18 days ago

Full Outage Map

Community Discussion

Tips? Frustrations? Share them here. Useful comments include a description of the problem, city and postal code.

Beware of "support numbers" or "recovery" accounts that might be posted below. Make sure to report and downvote those comments. Avoid posting your personal information.

Cloudflare Issues Reports

Latest outage, problems and issue reports in social media:

• 
  Jake (@JakeKing) reported 4 minutes ago

  Surprised to see that only 50% of internet traffic is now automated. the old "human good, bot bad" binary is dead. @Cloudflare scores every request 1-99 on behavioral trust instead.

• 
  Bo Montgomery (@BoBilbo28) reported 4 minutes ago

  @Dr_Crossroads I think this is a part of my thesis for investing in $NET. They are helping websites monetize the AI traffic that crawls their content. @eastdakota has talked about publishers and others working with Cloudflare to help them monetize their content with this move away from no clicks.

• 
  Sourav Debnath ✪ (@1axceler) reported 4 minutes ago

  @cagrisarigoz @Cloudflare Down 👎🏾

• 
  Stanley (@stanleyforx) reported 6 minutes ago

  Day 8 of doing 30 brutally honest account audits. @petecodes - 9,528 followers profile snapshot: - 38,924 total posts over 10.5 years (account since Dec 2014) - following 1,479 (6.4:1 ratio, decent but not dominant) - 125,854 likes given (extremely active engager) - verified blue check - pinned tweet: "best marketing I've spent so far" testimonial about High Signal newsletter ads. 5 likes, 2,579 views. weak pin - based in Edinburgh, UK - runs 3 things: ghostwriting service, High Signal newsletter, No CS Degree (dev inspiration) what's actually working: 1. strong opinions on tools people care about are his 67x multiplier. "If you don't like Webflow's price bump do this..." got 70,827 views, 235 likes, 11 RTs, 32 replies. his average post gets 1,070 views. this ONE post accounts for ~70% of all impressions in the entire dataset. when he picks a side on something trending and offers a specific alternative (Claude Code + Astro + Cloudflare), the algorithm goes all in. he's done this exactly once 2. questions that ask for personal experience. "How much are people paying for accountants in UK?" got 9 replies, 1,495 views. "How do you recover from bad sleep?" got 7 replies, 677 views. "Why do devs hate calls?" got 8 replies, 697 views. these consistently outperform his promos. people respond when asked about themselves 3. revenue transparency posts. "$5,950 in May 2026" got 13 likes, 1,183 views. monthly income reports consistently perform above his average. the indie hacker community rewards openness with real numbers 4. event and community content. the Nantes/Uneed Residency posts consistently hit 500-1,600+ views. real-life meetups create social proof and engagement from tagged people. "Great to see everyone at Uneed Residency" got 13 likes the problems: 1. newsletter promos are drowning the feed. ~30% of all posts are some variation of "subscribe to High Signal" or "buy a newsletter ad slot." most of these get 0-1 likes and under 150 views. the audience has completely tuned them out. the "High Signal" newsletter has become low-signal content on his feed 2. threads are structurally broken. the "If I was ghostwriting for @charlierward" thread: opener got 2 likes/163 views, parts 2-6 got 0 likes each and as low as 11 views. thread middles have no hooks, no tension, no reason to keep scrolling. he should either compress these into single posts or learn per-tweet hook structure 3. no consistent content pillar beyond self-promotion. remove the newsletter promos and ghostwriting pitches and there's no clear "why should I follow this person" content strategy. the Webflow post proves he CAN create massive value. but it's one post in 95 4. engagement rate is critically low. 0.60% average across the dataset. excluding the Webflow outlier, drops to ~0.35%. median post gets 270 views on 9.5K followers (2.8% reach). this suggests algorithmic suppression, likely from the high volume of low-engagement promo posts training the algorithm to not distribute his content 5. too many zero-value tweets. bare URL drops with no text, single-sentence non-sequiturs, reply-style tweets that read like they belong in someone else's thread. these dilute feed quality and signal to the algorithm that his content isn't worth distributing 6. the pinned tweet is a newsletter ad testimonial. first thing a new visitor sees is proof that buying ads in his newsletter works. that tells a potential FOLLOWER nothing about why they should follow. the Webflow post (235 likes, 70K views) should be pinned. it actually demonstrates his value 7. the bio is selling three things at once. ghostwriting service, newsletter, dev inspiration site. a new visitor has no idea what he's actually about. the identity is split three ways and none of them get enough oxygen wins he's leaving on the table: 1. the Webflow post format is his proven winner and he's never repeated it. "trending tool does something unpopular → here's the specific alternative stack I'd use instead." he could do this monthly. every time a SaaS raises prices, changes terms, or makes a controversial move, he has license to post "here's what I'd do instead" with a specific technical breakdown. he's done it once. it should be recurring 2. he's a ghostwriter who never shows his ghostwriting process. zero posts about how he analyzes a client's voice, how he structures a content calendar, what frameworks he uses. "here's how I'd rewrite this founder's last 5 tweets" would be the ultimate proof of his service. the ghostwriting pitch would sell itself through demonstration instead of self-promo 3. no CS Degree is a massive brand he doesn't leverage on X. he built an entire platform around developers without computer science degrees. that's a huge audience. but his X content barely mentions it. "this developer taught herself Python in 6 months and now makes $120K" stories from his own platform would be some of the most shareable content on dev Twitter 4. the revenue reports should be a monthly series with structure. "$5,950 in May" works. but "May 2026: $5,950. Here's what worked: [X]. What flopped: [Y]. What I'm trying in June: [Z]" would 3x the engagement because it gives people something to learn from, not just a number to see 5. he's an extremely active engager (125K likes given) but it's not translating. he's clearly reading and liking tons of content. but liking isn't engaging. thoughtful replies to larger accounts would convert that attention into followers. 10 replies per day to 50K+ accounts in the indie hacker space would move the needle more than 3 newsletter promos 6. the Edinburgh indie hacker angle is unused. building a solo business from Scotland, outside the SF/London bubble. that's a relatable narrative for thousands of remote builders. "building a $6K/month business from a cafe in Edinburgh" is a story. "subscribe to my newsletter" is not opportunities he hasn't seen: 1. "ghostwriter rewrites your tweet" series. take a follower's tweet (with permission), show the original, show his rewrite, explain the changes. this is his entire service demonstrated publicly. it sells the ghostwriting without ever asking for a sale. it's also insanely engaging because people love seeing their content featured and improved 2. "what I'd do differently if I started No CS Degree today" retrospective. he's been running it for years. the lessons from building a dev community without being a traditional CS grad himself would resonate with every self-taught developer on the platform. that's millions of people 3. leaning into the "I read 200 newsletters so you don't have to" curator angle. High Signal's value prop is curation. but his X promos just say "subscribe." instead, post the actual signal. "3 things I read this week that are worth your time" with genuine recommendations. give the value first, then the CTA earns its click 4. the "ghostwriting for [famous person]" format needs fixing, not killing. the concept is gold. "if I was ghostwriting for Elon, here's his content strategy in 5 tweets" is compelling. but the execution failed because the threads had no per-tweet hooks. compress each into a single post with the 3 biggest changes he'd make. one post, not a 6-part thread that dies at part 2 5. pricing and business model transparency. "what it actually costs to run a newsletter, a ghostwriting service, and a community site simultaneously." the indie hacker audience would devour this. real costs, real margins, what's worth it and what isn't. he's running 3 businesses. the meta-content about juggling them is more interesting than any of them individually 6. "here's what I learned ghostwriting for 10 different founders" synthesis post. no names needed. just patterns. "8 out of 10 founders make the same mistake in their first tweet of the day." that's a hook that drives both engagement and inbound leads for his service what he should double down on: 1. the contrarian tool take. proven at 67x his baseline. one per month when a SaaS does something unpopular. specific alternative stacks, not just complaints 2. show the ghostwriting work. public rewrites, before/afters, strategy breakdowns. sell the service by demonstrating it, never by promoting it 3. revenue transparency with lessons. monthly reports with what worked and what didn't. the number alone is a tweet. the number plus the lesson is content 4. kill the bare URL promos. every naked link with 0 likes is actively suppressing his reach. either add genuine value above the link or don't post it 5. compress threads into single posts. his thread middles die at 11-82 views. at 9.5K followers, one punchy post with the core insight will outperform a 6-part thread every single time bottom line: pete has 9.5K followers, three revenue streams, the proof that he can create a 70K-view post, and an extremely active engagement habit (125K likes). but 30% of his output is newsletter promos that get 0 likes, his threads structurally collapse after tweet 1, and his ghostwriting expertise is completely invisible on his feed. the fix is simple: stop promoting and start demonstrating. show the ghostwriting process, give the newsletter value before asking for the subscribe, and repeat the Webflow format that already proved it works at 67x his baseline. what do you say @petecodes, did i get it right? hit me up if you want to brainstorm more ideas together.

• 
  Elijah Onato (@ElijahOnato) reported 7 minutes ago

  @NordEye I remember seeing this photo when cloudflare servers went down

• 
  Mr. Review Ai (@MrReviewai) reported 8 minutes ago

  3/ The Fix Strategy for WordPress: No hiding the ugly numbers. I’m stripping down unused CSS/JS using Asset CleanUp, setting up advanced caching, and testing Cloudflare routing to optimize global delivery.

• 
  Agasta (@idkAgasta) reported 14 minutes ago

  @anirudhprmar But you will lose global cdn/ edge network with this .... Also I don't like to host forntend or full stacks in a single vm..... I will consider using Cloudflare workers if the budget is tight or AWS cloudfront + ecs with fargate if we are richy-rich

• 
  Taylor Page (@TRPage_dev) reported 15 minutes ago

  We complain a lot about Shopify Support, but I don't think I can anymore. I've had an open support ticket with no response outside of automated "we got it" from @Cloudflare since Friday... Turns out we're still ahead of the curve.

• 
  Dearg OBartuin (@dearg_x) reported 16 minutes ago

  Do you ever feel like you personally broke the internet? Adding a new domain to @Cloudflare next of all ...... global outage - 💥

• 
  Joe Can Write (@joecanwrite) reported 17 minutes ago

  Bots are now 57.3% of all requests to web pages, per Cloudflare. Human traffic is the minority. If you judge content on raw pageviews, a growing share of your audience is machines that never buy anything. The question that matters now: do AI systems recommend you?

• 
  Aman (@aman_bagrecha) reported 18 minutes ago

  @spatialthoughts It was! It said it is possible due to me having a password login (they brute forced into guessing password). Two fixes I did: have an identity key (ssh admission) rather than password. And then router my traffic through cloudflare so hacker cannot know my direct IP

• 
  Aryan Esfandiari (@arian88) reported 21 minutes ago

  Never push to main on a Friday @Cloudflare

• 
  MoroJS (@Moro_Js) reported 27 minutes ago

  @samgoodwin89 cloudflare alchemy is cool but if you’re building the api layer, why not skip the boilerplate? morojs ships with cloudflare workers support, built-in caching, and auto-typed routes. same code, 10ms cold starts. no config files. just `createApp()` and ship. 🚀

• 
  Leon (@Nordikkkk) reported 27 minutes ago

  @PlutoPurityGG @nthglsn @Cloudflare They did receive the service, you pay retroactively

• 
  Z (@Cosmosz) reported 28 minutes ago

  is Cloudflare down? can't load account

• 
  Hasan Aboul Hasan (@hasan_ab_hasan) reported 30 minutes ago

  Day 2 of building ToolerBox: the largest free online tools site on the web. Today I added country analytics to the admin dashboard. Every login now records which country it came from. If an account later logs in from a new country, it gets flagged. I thought this needed a paid geolocation service. It didn't. Why I built it: A new-country login is one of the cheapest signals an account is shared or stolen. And knowing where users actually are guides what to localize and build next. How it works: Cloudflare already geolocates every visitor at the edge and stamps a country code on requests. A free header, CF-IPCountry. No MaxMind database, no geolocation API, no latency. But there's a catch. A header is just text the client sends. Anyone can hit my server directly and send "CF-IPCountry: US". So NEVER trust it. The fix: only trust the header when the IP connecting to you is provably Cloudflare. A forged header sent to the origin gets ignored, its IP isn't Cloudflare's. Trust the header, but only from a source that can't fake it. A few other decisions: - Store the country on the user row, so the admin list stays fast, no per-user joins. - Increment the per-country counter atomically, so two logins at once don't lose a count. That's a race condition. - Fire the alert ONCE, on the first mismatch, not on every foreign login. It's a review signal, not an auto-block. VPNs, proxies, and travel exist. The result: shared or stolen accounts land on a reviewable list, plus real location data, no analytics vendor. And yes, I know this looks like overkill on day 2. But ToolerBox is my lab to learn and build real, scalable, production-ready apps. I'd rather build it now and grow into it than bolt it on later.

• 
  Sarthak Rawool (@sarthakcore) reported 31 minutes ago

  my tool graveyard: Notion for project management. too complex. back to markdown. Obsidian + Hermes for a knowledge vault. never found the use case. Super X for managing my account. replaced it. Codex for writing code. now it only runs test passes. every tool i abandoned taught me the same thing: if it doesn't fit how i actually work, it dies. current stack that survived: Claude Code 6+ hours/day. Claude Design for UI. PlanetScale. Dodo Payments. Cloudflare + GitHub Actions. the tools that stick are the ones i forget i'm using.

• 
  Extagonist (@extagonist) reported 32 minutes ago

  @adamlyttleapps @SynergyWS Just get a droplet or any VPS and use a free cloudflare tunnel and lock down the server ports for everything aside from port 22 for SSH

• 
  Youssef 🚜 (@yelkhayami) reported 34 minutes ago

  we're a small team running 5 consumer apps, so checking posthog, cloudflare, search console and slack for every app every day stopped being realistic i ended up pulling everything into one internal tool, one timeline per app: • error logs (client + server) • analytics events we care about • cloudflare logs • google search console issues • relevant slack channels • custom signals per app, orders, signups, queue sizes, anything a db query can answer agents read whatever is new every 30 minutes and only surface findings that need review. failed crons, seo drops, weird spikes. i almost NEVER reads raw logs anymore it's also an mcp server, so claude can query the whole company when we're debugging or making decisions, so my team can use it too we never set out to build a 'data foundation', it grew out of not wanting to check 5 dashboards but now that we have it up and running it's honestly great

• 
  TBPN (@tbpn) reported 35 minutes ago

  FULL INTERVIEW: Cloudflare CEO @eastdakota joins TBPN to discuss why agent traffic has surpassed human web traffic, the company's acquisition of VoidZero, and why concerns about data center water usage are overblown. 2:13 - Why today's infrastructure can't support billions of AI agents 7:38 - Bot traffic vs. human traffic 11:51 - Long-running AI agents are the future, not chatbots 15:15 - The 3 reasons companies are moving AI inference to the edge 17:29 - On concerns about data centers using too much water 19:41 - Matthew Prince on lawsuits from Spain and Italy over piracy 22:10 - Why being a public company is healthier than taking VC money 28:57 - Matthew Prince on hiring 1,111 interns 2:13 - Why today's infrastructure can't support billions of AI agents 7:38 - Bot traffic vs. human traffic 11:51 - Long-running AI agents are the future, not chatbots 15:15 - The 3 reasons companies are moving AI inference to the edge 17:29 - On concerns about data centers using too much water 19:41 - Matthew Prince on lawsuits from Spain and Italy over piracy 22:10 - Why being a public company is healthier than taking VC money 28:57 - Matthew Prince on hiring 1,111 interns

• 
  Sussycat Bloomberg (@LeSussyCat) reported 38 minutes ago

  @Cloudflare Put a smile on my face because atleast DDoS is using my ****

• 
  David Siewert (@david1gp) reported 38 minutes ago

  Cloudflare Pages not working? Or I the only one who is affected? ```json {"text":"POST /pages/assets/upload -> 502 Bad Gateway"},{"text":"Cloudflare Ray ID: a0a71049ef6c7049-FRU"}],"kind":"error","name":"APIError"} ``` @CloudflareDev

• 
  Daniel Romero (@HyperTechInvest) reported 43 minutes ago

  Cloudflare $NET believes CPU demand could 20x with AI agents $AMD $INTC $ARM The company argues that AI agents could create a massive CPU/server infrastructure problem if every worker runs multiple agents using today’s cloud model The math: US: ⏩ 100M knowledge workers × 1 agent each ÷ ~10 agents per CPU = 10M CPUs Global: ⏩ 1B knowledge workers × 10 agents each ÷ ~10 agents per CPU = 1B CPUs Cloudflare compares this to current global server CPU production of only ~35M–45M per year, implying the current approach could require ~20x current annual CPU production $NET is pitching Agent Cloud as a solution This is how it would work: 1. Dynamic Workers instead of full containers Agents can be spun up on demand and put back into cold storage when idle, avoiding one heavy always-on container per agent 2. Durable Objects for state Agents need memory/state for tasks, files, sessions, progress, and tool outputs. Durable Objects give each agent/app a stateful object with local SQLite-backed storage 3. Workflows for long-running agents Agents may run for minutes, hours, or days while waiting for approvals, retries, external events, or scheduled steps. Dynamic Workflows allow these processes to hibernate between steps instead of constantly consuming compute 4. Project Think / Agents SDK Project Think adds agent-specific primitives like durable execution, crash recovery, checkpointing, sub-agents, persistent sessions, and sandboxed code execution 5. Sandboxes only when full computers are needed For coding agents or tasks requiring bash, ***, file systems, browsers, or arbitrary binaries, Cloudflare still offers Sandbox containers. But the idea is to use them only when necessary, not as the default runtime for every agent

• 
  aff (@affoehteimoso) reported 46 minutes ago

  is cloudflare down?

• 
  Dhairya (@dkare1009) reported 46 minutes ago

  📂 SaaS Stack ┃ ┣ 📂 Frontend ┃ ┣ 📂 React ┃ ┣ 📂 NextJS ┃ ┣ 📂 Vue ┃ ┣ 📂 TailwindCSS ┃ ┗ 📂 Shadcn UI ┃ ┣ 📂 Backend ┃ ┣ 📂 NodeJS ┃ ┣ 📂 Django ┃ ┣ 📂 Laravel ┃ ┣ 📂 FastAPI ┃ ┗ 📂 Express ┃ ┣ 📂 Database ┃ ┣ 📂 PostgreSQL ┃ ┣ 📂 MySQL ┃ ┣ 📂 MongoDB ┃ ┣ 📂 Redis ┃ ┗ 📂 Supabase ┃ ┣ 📂 Auth ┃ ┣ 📂 Clerk ┃ ┣ 📂 Auth0 ┃ ┣ 📂 Firebase Auth ┃ ┣ 📂 Supabase Auth ┃ ┗ 📂 NextAuth ┃ ┣ 📂 Payments ┃ ┣ 📂 Stripe ┃ ┣ 📂 Paddle ┃ ┣ 📂 Dodo Payments ┃ ┣ 📂 Lemon Squeezy ┃ ┗ 📂 Polar ┃ ┣ 📂 Emails ┃ ┣ 📂 Resend ┃ ┣ 📂 SendGrid ┃ ┣ 📂 Mailgun ┃ ┣ 📂 Postmark ┃ ┗ 📂 Amazon SES ┃ ┣ 📂 Storage ┃ ┣ 📂 AWS ┃ ┣ 📂 Cloudflare ┃ ┣ 📂 Google Cloud Storage ┃ ┣ 📂 Supabase Storage ┃ ┗ 📂 Uploadcare ┃ ┣ 📂 Deployment ┃ ┣ 📂 Vercel ┃ ┣ 📂 Netlify ┃ ┣ 📂 Railway ┃ ┣ 📂 Render ┃ ┗ 📂 AWS ┃ ┣ 📂 Domains and DNS ┃ ┣ 📂 Namecheap ┃ ┣ 📂 Hostinger ┃ ┣ 📂 Cloudflare DNS ┃ ┣ 📂 Google Domains ┃ ┗ 📂 SiteGround ┃ ┣ 📂 Analytics ┃ ┣ 📂 Google Analytics ┃ ┣ 📂 Plausible ┃ ┣ 📂 PostHog ┃ ┣ 📂 Mixpanel ┃ ┗ 📂 DataFast ┃ ┣ 📂 Monitoring ┃ ┣ 📂 Sentry ┃ ┣ 📂 LogRocket ┃ ┣ 📂 Datadog ┃ ┣ 📂 NewRelic ┃ ┗ 📂 UptimeRobot ┃ ┣ 📂 DevOps ┃ ┣ 📂 Docker ┃ ┣ 📂 Kubernetes ┃ ┣ 📂 GitHub Actions ┃ ┣ 📂 CI CD ┃ ┗ 📂 Terraform ┃ ┣ 📂 Search ┃ ┣ 📂 Algolia ┃ ┣ 📂 Meilisearch ┃ ┣ 📂 Elasticsearch ┃ ┣ 📂 Typesense ┃ ┗ 📂 OpenSearch ┃ ┣ 📂 AI Integration ┃ ┣ 📂 OpenAI API ┃ ┣ 📂 Anthropic API ┃ ┣ 📂 Replicate ┃ ┣ 📂 HuggingFace ┃ ┗ 📂 Gemini API ┃ ┣ 📂 Integrations ┃ ┣ 📂 Zapier ┃ ┣ 📂 Make ┃ ┣ 📂 n8n ┃ ┣ 📂 Pabbly ┃ ┗ 📂 Webhooks ┃ ┣ 📂 Security ┃ ┣ 📂 SSL ┃ ┣ 📂 Cloudflare ┃ ┣ 📂 WAF ┃ ┣ 📂 Rate Limiting ┃ ┗ 📂 Secrets Management ┃ ┣ 📂 Marketing ┃ ┣ 📂 Search Console ┃ ┣ 📂 Outrank ┃ ┣ 📂 Buffer ┃ ┣ 📂 Analytics ┃ ┗ 📂 Kit ┃ ┗ 📂 Customer Support ┣ 📂 Intercom ┣ 📂 Crisp ┣ 📂 Zendesk ┣ 📂 Tawk ┗ 📂 HelpScout

• 
  sam (@samgoodwin89) reported 47 minutes ago

  A beautiful "software factory" with its own "software byproducts". As Fable generates 100% Cloudflare IaC coverage, it also produces a perfectly patched API spec and Effect SDK. All important errors and fixed data types are discovered from the API's real behavior (PR below)

• 
  Sadiq (zk arc) (@Md_Sadiq_Md) reported 52 minutes ago

  @0xRasmPro @Cloudflare @tan_stack Quartz solves 95% of the problems, but the math renders took a ton of time for me to solve

• 
  Jonathan Moore (@Moore) reported 54 minutes ago

  @TRPage_dev @Cloudflare Funny… I opened a Shopify support ticket and they were able to quickly confirm the CDN issue we had was coming from a Cloudflare outage.

• 
  Maximum YT (@Maximum__YT) reported 57 minutes ago

  @status_is_down This seems like a Canada wide issue, cloudflare and telus is down.

• 
  Stealth Exploit (@stealthexploit) reported 60 minutes ago

  Sorry guys I was not available yesterday so here is a summary Google Cloud Run (Hosting) Cloud Run only charges you for the exact milliseconds your app is processing a request. If nobody is visiting the site, the servers instantly scale to zero and you pay absolutely nothing. With Cloud Run you get 2 Million free requests, 360,000 GB-seconds of memory, and 180,000 vCPU-seconds completely free every single month. Cost: $0.00 until you start getting thousands of daily active users Artifact Registry (Storage) This is usually where companies bleed money because old, massive Docker images pile up over months. The Free Tier gives you 0.5 GB of free storage per month but i created the policy.json to aggressively delete images older than 7 days and only keep the last 3 versions this would make my storage incredibly small. Our Nginx/Alpine containers are extremely lightweight (~25MB each). Cost: $0.00 to $0.10 per month. Google Cloud Build (CI/CD) Every time you push code, Google spins up a machine to build your Docker image. The Free Tier also gives you get 120 free build-minutes every single day but since i was using Vite, builds take about 1 to 2 minutes. This means me and my team can push code 60+ times a day for free with an estimated cost of $0.00 Network Bandwidth (Egress) Sending data to users costs moneyso google gives you 200GB per month for free, For better optimization use Cloudflare in front of your domains, Cloudflare acts as a massive shield and It aggressively caches your static CSS, JS, and Images on their own edge servers around the world for free. This means only a tiny fraction of your traffic ever actually hits Google Cloud. Estimated Cost: $0.00 cc : @maazscript @LanHubs