Satellite Shorts July 2026

Accelerated Satellite Regulatory Approvals. The FCC announced that it will vote at its July 22 meeting to overhaul the regulatory process for satellite broadband. The proposed rules are in an order titled Space Modernization for the 21st Century. The proposed rules are to meet four specific goals: to improve the speed of processing applications for new satellites, to provide more predictability to applicants and licensees, to provide more flexibility for innovation, and to still meet the FCC’s responsibilities. It’s one of the longest FCC orders I can recall, at 296 pages.

The new rules would eliminate a lot of old regulations and replace them with new rules intended to speed up the paperwork process by applying “bright-line” criteria for approving new satellites. For those not familiar with that term, a bright-line regulation is one with a clear, objective rule or standard that leaves no room for subjective interpretation or exceptions, that is designed to create highly predictable outcome. A bright-line approval process would list specific requirements, and any applicant who meets those requirements would be approved. Perhaps I’m naïve, but I would expect that approving new satellites should include a thorough engineering review at the FCC. There should also be an opportunity for other satellite providers to note any problems with an application. I’m all for improving the processing time at regulatory agencies, but changing from today’s through review to a set of bright-line rules is going too far in the direction of simplification and assumes that satellite companies won’t propose any bad or unworkable ideas.

100,000 Satellites for Starlink? Perhaps this is a coincidence for the FCC wanting to simplify satellite applications, but Starlink filed an application to expand to 100,000 broadband satellites to its constellation, up from the 12,000 that will be in orbit at the end of this year and the 15,000 total that have been authorized by the FCC. The application includes a request to get access to new spectrum in the W and D bands between 92 and 275 GHz. The filing says that the new satellites will be the new Gen3 that will be capable of speeds up to several gigabits per second. Never one for modesty, the filing from Elon Musk says that SpaceX has a goal of handling the majority of the world’s internet traffic.

Rocket Lab Buys Iridium. In an $8 billion deal, Rocket Lab is acquiring the satellite company Iridium. Iridium currently operates a constellation of 66 satellites that provide a range of services that include phone and data services using the L-band spectrum, positioning services, and navigation and timing services. Rocket Labs is a launch service that hopes the deal will make the company into a full-service provider that both launches and operates satellites. The company hinted earlier this year that it was interested in developing its own constellation but said it’s not ready to announce specific plans and services until it acquires the needed spectrum.

NASA Warns About Launch Site Capacity. A report from the inspector general of NASA warns that the increasing number of space launches is posing problems for spaceports. The report warns that the Kennedy Space Center in Florida and the Wallops Flight Facility in Virginia are approaching operational capacity and will hit their limits between 2028 and 2029. The prediction is that the number of launches will grow from the current 109 annual launches to 268 per year by 2030, which will overstrain the launch facilities that were constructed years ago to support the Apollo program. The report warns that an investment of at least $1 billion is needed to future-proof the sites. The launch sites are controlled under the NASA budget, and the administration has proposed a 24% cut in NASA’s budget for the next fiscal year, which includes eliminating forty scientific programs.

Premium Fees from Starlink? A recent article in Futurism reports that SpaceX is charging some users a demand surcharge of $500 to $1,500 because their address is within a high-demand area. The article cites several customers who have been hit with the surcharge, and who complain that it’s nearly impossible to fight the issue through the company’s slim customer service process.

The article notes that the congestion charge was quietly reintroduced in 2024 at a one-time fee of $100 for high-congestion areas. PC Magazine reported in 2025 that the fees had been increased to as much as $1,000. As of last month, there are reports of surcharges of $1,500 in parts of Alaska.

Meanwhile, Starlink has increased prices. While it now offers usage-capped plans for $55 or $85 in selected areas, the price for normal unlimited Starlink broadband was increased from $120 to $130.

Behind the Scenes at T-Mobile

There is some interesting corporate maneuvering happening behind the scenes at T-Mobile. A recent article in the Wall Street Journal (behind a paywall) talks about the plans that Timotheus Höttges, the CEO of Deutsche Telekom, has for his company. The company is already the biggest cellular company in the world with over 273 million mobile customers in fifty countries.

Höttges is now trying to orchestrate a full merger between the two firms. Deutsche Telekom currently owns 54% of T-Mobile, and during his twelve years in charge of the company, he’s changed T-Mobile from a company that perpetually lost money to one of the most recognizable brand names in the industry. He thinks a merger is needed to give T-Mobile the resources it needs to fully succeed.

T-Mobile recognizes that the key to success is to own a large number of both fiber and cellular customers. The company has been engaging in partnership deals to gain a share of fiber overbuilders. A few years ago, the company purchased a 50% stake in Metronet, a Midwest fiber overbuilder. The company engaged in a joint venture with EQT to acquire Lumos. More recently, T-Mobile purchased 50% of i3 Broadband, GoNetspeed, and Greenlight Networks. The company bought all of the ownership of US Internet in the Twin Cities.

The Wall Street Journal reports that T-Mobile is financially constrained from growing through big acquisitions and deals. The current T-Mobile corporate structure makes it impossible for the company to buy companies by issuing stocks without diluting the ownership of Deutsche Telekom, and it’s limited on how much debt it can take on.

Höttges believes that combining Deutsche Telekom and T-Mobile into a giant $300 billion company will allow T-Mobile to take on the debt needed to grow. He thinks T-Mobile should be competing with AT&T and Verizon, and that’s going to take big acquisitions.

Höttges apparently has his work cut out for him. Current T-Mobile shareholders might not be interested in gaining ownership of the lower-margin Deutsche Telekom. He also needs to convince the German government, which owns a 28% share of Deutsche Telekom. And if he can pull off those steps, he has a lot of work to do to gain regulatory approval in the U.S. and Europe.

You have to wonder where an unconstrained T-Mobile would look for growth. There are still additional mid-sized and smaller fiber overbuilders it could pursue. The largest cable company that could be on the market is Altice, which is mired in a lawsuit against Apollo Global Management, Ares Management, and BlackRock that accuses the companies of trying to force it into bankruptcy.

The only other large companies that might make sense for a merger with T-Mobile would be Charter or Comcast. I’ve been reading several analysts lately who think the big blockbuster mergers are inevitable. But some think a more natural suitor for these companies would be SpaceX.

There is a lot that has to happen for Deutsche Telekom and T-Mobile before any big blockbuster deal can be contemplated.  For those of us who enjoy watching the big boys maneuver, the next few years are going to be very interesting.

Is AI Changing Traffic Patterns?

Mitch Wagner of Fierce Network recently published an article that claims that AI is blowing up 30 years of traffic network assumptions. He claims that AI traffic is smoothing the daily peaks and dips in network traffic that all ISPs are familiar with. While every ISP is a little different, any ISP that serves a lot of end-user customers expects traffic peaks in the evening, smaller peaks during the daytime, and very low levels of network traffic at night.

Network engineers have always paid close attention to the peaks, which were the main factor in determining the size of network connections. Nobody wants to have a network that restricts bandwidth when customers want to use it the most. The cable companies learned this lesson the hard way during the pandemic when customers suddenly needed to work and school from home and found the broadband connections unable to meet their needs, particularly in homes where more than one person wanted to use the network at the same time. Every network engineer I know can cite the busy hour, busy day, and busy week on the networks they manage.

Wager says the peaks in traffic are evening out and that networks are seeing a more consistent demand throughout the day. Wagner cites Ed Fox, the CTO of MetTel, from an interview given for a Fierce Network Research report.

I have no reason not to believe Mr. Fox. However, MetTel is an ISP that operates in major urban centers and likely serves the kinds of businesses that have become heavy AI users. It’s not hard to imagine that an urban ISP serving businesses might be seeing a drastic change in traditional network traffic patterns. But I have to think that MetTel and other urban business-centric ISPs didn’t have the same traffic patterns as other ISPs before the advent of AI.

I work with a number of ISPs and I have not heard anybody talking about a big change in traffic patterns. My clients work in a variety of markets, from rural to urban, but none concentrate on the urban market business that MetTel is experiencing in places like New York City.

However, I appreciate the article, because if it’s true, then most ISPs, except for fully rural ones, might eventually see some shift in traffic patterns due to AI. I am curious to get feedback from this blog from IPS to hear if anybody is seeing anything like the big changes being experienced by MetTel.

The Wagner article also claims that upstream bandwidth is growing faster than downstream, reversing a 30-year design assumption built around heavy downloads. This is something that has been well documented by OpenVault. They’ve been shown that upstream usage has been growing faster than download usage starting in 2024. In the first quarter of 2024, national average upload usage increased by 13.2% compared to 7.7% for download. In the first quarter of this year, average upload growth was at 19.8% while download growth was at 7.8%. OpenVault credits most of the growth in upload usage to computer software synching with data centers.

Wagner claims the upsurge in upstream usage comes from AI inference traffic moving towards the edge, particularly for video processing. What he means by that is big growth coming from uses of video cameras for functions like AI-driven video surveillance at retail locations, camera-equipped wearables, and cloud-based operational technology in applications like oil and gas asset management. I hope that the big national companies that monitor traffic begin tracking this issue. I’d love to hear more about the trends in specific traffic, like video surveillance and wearables.

Network engineers all understand that upload traffic is usually a tiny fraction of download usage, with average download usage a dozen times more than upload. With the possible exception of cable companies, upload usage is largely an afterthought for network engineers, who barely consider it when sizing and designing networks.

I have no doubts that there are localized situations where AI traffic is making a big difference in network traffic. But for ISPs that mostly serve residential and small business customers, I still have to think it’s a tiny, possibly unnoticeable blip.

The Future of State Broadband Offices

Kathryn de Wit and Jake Varn of Pew recently wrote an article that cautioned that States Must Consider Future of Broadband Offices. They note that some states have a sunset date embedded in the enabling legislation that will mean the end of State Broadband Offices if state legislatures don’t act.

I’ve also been thinking about this lately. States have been busy in recent years overseeing broadband grants that were funded by the Capital Project Fund. Many legislatures augmented those grants with funding from ARPA for additional grant funding. Both of those programs are finished this year, other than a few waivers to extend funding until July of next year. Before these two programs, the States oversaw the use of CARES Act funding, which was used for a wide variety of purposes. A handful of legislatures also funded broadband grants out of the state coffers.

Most States are busy right now trying to get BEAD grants in place with ISPs, although a few States have only minuscule outlays for BEAD. For unknown reasons, NTIA has still not agreed to the grants for Illinois and California. BEAD grant construction is supposed to conclude in four years, and States are on the hook to verify that construction meets the BEAD specifications.

There were two other sources of federal funding that were supposed to feed through the States. The biggest source is the BEAD non-deployment funds, which is whatever is left over from the $42.5 billion grant program after funding infrastructure. This was originally a relatively small amount, and in 2024, thirteen States told NTIA that they would probably have excess funding left over after infrastructure grants. However, when NTIA implemented the Benefit of the Bargain rules, it slashed infrastructure grants, and non-deployment has ballooned to over $21 billion. There is still no assurance that this money will ever be given to States to spend. NTIA has repeatedly pushed off the date when it will disclose the use of these funds. The U.S. Treasury continues to refer to these funds as net savings, implying they won’t be spent.

The other funding that is still up in the air is the $2.75 billion in grants that were funded by the Digital Equity Act. A lot of this funding was to be administered by the States to provide computers and training to take advantage of the new infrastructure being built through federal grants. The administration abruptly canceled this grant program, supposedly because it included the word equity in the title. The National Digital Inclusion Alliance (NDIA) sued the administration over the end of the grant. It’s possible that this funding still has legs. The DOJ told a court last week that the government might remove its objections to the grant program if all references to race were removed. But there is probably still a long road to seeing this funding since the administration’s proposed budget for next year eliminates these funds.

I have to wonder what happens to State Broadband Offices if non-deployment or Digital Equity Grants are never funded. They will be left overseeing the invoices for BEAD. States are on the hook to measure speeds on grant programs for another decade – but will that be enough to convince legislatures to keep funding broadband offices?

Some States will fund new grant programs to bring infrastructure to the places missed by BEAD. Wisconsin has already announced a new state grant program, and there will probably be another half dozen states that issue grants to continue to close the broadband gap. But a lot of States now believe they are mostly covered with decent broadband, and there won’t be any incentive for those States to continue to pay for a broadband office.

Pew suggests activities that broadband offices should pursue after BEAD. This includes setting broadband goals, collecting and mapping broadband data, and providing technical assistance to communities and stakeholders. I would extend that list to suggest that States focus on digital inclusion and affordability efforts to make sure that everybody can actually use the broadband networks that have been built.

But I have to wonder how much traction these goals will generate in States that are looking at meager and tight overall budgets in the foreseeable future. Is broadband going to remain enough of a priority at State legislatures to attract funding when so many other important functions are losing federal funding? Will States that think broadband has been solved care about maintaining a broadband grant office? We have to remember that, before the CARES Act many States did not have a formal broadband office. I suspect that when the sunset date hits for broadband offices, that some States will let the function lapse in favor of other priorities. I can’t see the States caring very much about NTIA requirements to measure broadband speeds for BEAD networks once those networks are funded and operational.

Why Are We Building 6G?

I was recently reading the preliminary specifications for 6G, which led me to wonder why the cellular industry is willing to make a huge investment in a new technology that is likely not going to drive a lot of new revenue opportunities. I’m admittedly not a wireless guy, so perhaps I’m missing something. But consider the claims being made for 6G.

Faster Speeds. 6G will support peak data rates at speeds between 50 and 100 Gbps. Is there really a market for faster cellular speeds? Are cellular customers crying out for faster speeds than can be supported by 5G (supposedly 10 Gbps)? What’s not said is that the highest speeds touted by the specification can only come through deployment of superhigh spectrum, which also comes with severe distance limitations outdoors.

Lower Latency. The 6G spec targets latency between 0.1 and 1 milliseconds, which is better than the 4 millisecond goal of 5G. But are there many customers willing to pay extra for that level of low latency?

Enables Immersive Communications. 6G is said to enable technologies like immersive eXtended Reality (XR), remote multi-sensory telepresence, holographic communications, haptic sensors and actuators, and multi-sensory interfaces. While these all sound supercool, and will certainly have proponents, is there enough revenue opportunity from these technologies to justify a massive upgrade of cell towers? Over the last five years, we’ve seen Meta completely fail to convince the world to use the metaverse using fiber connections. Technologies like smart glasses have been a total flop.

These technologies all require significant bandwidth, implying that we’d have to bring cell towers closer to users. Are cellular carriers suddenly going to invest in the small cell network that largely never materialized with 5G? Cellular carriers love the advantages of small cells, but they don’t want to pay for the fiber connectivity needed to support towers everywhere.

Connections for Smart Machines. This means things like autonomous factories and robots. 5G promised smart factories, and there have been some built – but there was not nearly enough revenue from these to support the huge investments in ubiquitous 5G. It’s a very different technology to broadcast 5G inside a factory than from every cell tower. Maybe I have a faulty crystal ball – does anybody see the country swarming with robots in the next decade?

Ability to Connect to Massive Numbers of Devices. This was the big hope for 5G. Cell carriers thought that every car would have a separate cellular subscription. They assumed that farms would be using massive numbers of cellular sensors. They assumed that our homes would be filled with smart appliances and devices connected to 5G.

But this didn’t happen. Iain Morris, the International Editor of Light Reading, recently published an article that cautioned that cellular carriers will be repeating their past commercial failures if they pin their hopes on 6G capturing big revenues from IoT.

The article reminded me of some of the early predictions related to IoT. Dave Evans of Cisco predicted in 2015 that the world would have 50 billion IoT devices by 2020. According to the recent Ericsson Mobility Report, the world had 22.3 IoT devices at the end of 2025, massively short of the 2015 prediction. It’s still an impressive number, since it means there are now almost three IoT devices for every person on earth. But as Morris points out, very few of these devices use 5G. WiFi and other technologies like Loran have captured the IoT market.

The issue that killed the IoT market for 5G hasn’t changed for 6G. It’s still far too costly to use cellular technology to monitor devices that only periodically connect to the Internet. And that is before considering the cost of a cellular subscription.

What’s most troublesome to me about the advent of 6G is that it will be accompanied by the sunsetting of 4G. In the U.S., that is going to mean that rural cellular coverage is going to shrink to a tiny percentage of what is covered today as I discussed in a recent series of blogs. 5G has proven to be a wonderful urban technology. 5G was a mandatory update for cellular companies since 4G was unable to support the increasing numbers of urban cellphones, particularly when customers increased the use of cellular data. But I’m having a hard time seeing the same need for the expensive upgrade to 6G. I know cellular carriers must be quietly having this same conversation.

AT&T v. Duke Energy

On June 24, AT&T filed a complaint with the FCC against Duke Energy Carolinas about the rates being charged for pole attachments. AT&T alleges that Duke is charging rates far higher than allowed by law in North and South Carolina. AT&T claims it is entitled to pay “just and reasonable” rates under FCC rules. AT&T is asking that Duke be required to refund overcharges.

It’s an interesting complaint for several reasons. This is more of a partnership complaint than a straight complaint about how pole attachment rates are calculated. AT&T and Duke entered into a Joint Use Agreement (JUA) in 1978 since they share ownership of poles in the region. There are 457,901 poles covered by the JUA, with Duke owning 80% of the poles and the remaining 20% owned by AT&T.

One of the nuances of the case is jurisdiction over pole attachment regulation. North Carolina exercised  reverse-preemption of pole attachments, while South Carolina remains with FCC regulations concerning poles. AT&T claims the FCC has authority over the dispute in both states. I have to wonder why this agreement doesn’t fall partially under North Carolina’s jurisdiction, and if the state has somehow conceded authority to the FCC.

The dispute centers around FCC rules included in 47 U.S.C. § 224 that determine the maximum rate that can be charged for a pole attachment. Those rules calculate two different pole attachment rates, one that applies to cable companies and a telecom rate that applies to everyone else. The maximum rate for telecommunications carriers is designed to ensure that a telecom provider pays a proportional share of both the usable space and unusable space on the pole, divided by the total number of attaching entities. There are specific formulas defined by the FCC for the rate calculation, and calculating the maximum rate is mostly an exercise in gathering the right accounting data to populate the formula.

AT&T alleges that Duke is charging them a rate far in excess of the maximum allowed telecom rate. We can only guess how much higher since the public version of this complaint has redacted the higher rates. I can’t imagine how the public would be harmed by knowing the higher rates. There is a table in the complaint that calculates the average telecom rates from 2023 through 2026 at $10.92.

I characterized this earlier as a partnership dispute. AT&T is complaining that the rates that AT&T and Duke pay under the JUA are disproportionate to the amount of space each uses on the poles, such that AT&T pays far more than Duke on a per-foot basis. AT&T also complains that Duke benefits by being able to offset its costs by fees charged to other attachers, something not available to AT&T. Both of these sound more like complaints related to the old Joint Use Agreement that don’t seem relevant to FCC regulation.

AT&T’s argument is largely based on an FCC Order from 2018 that said that the telecom pole attachment rate applies to all new and newly-renewed joint use agreements, including agreements that are automatically renewed or extended.

You may be asking, if you read this far, why I chose to write about this dispute. There are several reasons. First, it’s always fascinating to get a glimpse behind the curtains of the deals made between big companies that we would otherwise never know about. I’m sure that a fiber builder asking to get on a Duke pole in North Carolina has no idea that the poles in question might actually be owned by AT&T. It’s also interesting to see how a big power company like Duke might decide to overcharge a partner in the pole business. The Joint Use Agreement was reached in 1978, which is ancient history in the corporate world. Somebody at Duke probably got a bonus for increasing pole attachment fees to AT&T in violation of an old agreement they might not even been aware of. Finally, I’m a customer of both Duke and AT&T, and regardless of how it resolves, I’m betting that I won’t see any benefit from the decision. This kind of dispute affects stockholder profits and not rates charged to the public.

For the Holiday

For the holiday, I’m republishing my most popular blog that talks about critters and broadband networks. Enjoy your weekend and stay cool!

Most people don’t realize the damage done every year to fiber and other wired networks by animals.

Squirrels. These cute rodents are the number one culprit for animal damage to aerial fiber. To a lesser degree, fiber owners report similar damage by rats and mice. Squirrels mainly chew on cables as a way to sharpen their teeth. Squirrel teeth grow up to 8 inches per year, and if squirrels aren’t wearing their teeth down from their diet, they look for other things to chew. There has been speculation that squirrels prefer fiber to other cables due to some oil or compound used in the fiber manufacturing process that attracts them.

Level 3, before it was part of CenturyLink, reported that 17% of aerial fiber outages were caused by squirrels. A Google search turns up numerous outages caused by squirrels.

Companies use a wide variety of techniques to try to protect from squirrel damage – but anybody that has ever put out a bird feeder knows how persistent they can be. One deterrent is to use hardened cables that are a challenge for squirrels to chew through. However, there have been cases reported where squirrels partially chew through such cables, which still lets in water and can cause future damage.

A more common solution is adding a barrier to keep squirrels away from the cable. There are barrier devices that can be mounted on the pole to block squirrels from moving higher. There are also barriers that are mounted where cables meet a pole to keep the squirrels away from the fiber. There are companies that have tried more exotic solutions, like deploying ultrasonic blasters to drive squirrels away from fiber. In other countries, the fiber providers sometimes deploy poisonous or obnoxious chemicals to keep squirrels away from the fiber. These techniques are frowned upon or illegal in the US.

Gophers. Buried fiber also has a gnawing pest in the pocket gopher. There are thirteen species of pocket gophers that range from 5 to 13 inches in length. The two parts of the country with the most pocket gophers are the Midwest plains and the Southwest. Gophers live on plants and either eat roots or pull plants down through the soil.

Pocket gophers can cause considerable damage to fiber. These rodents will chew almost anything, and there have been reported outages from gophers that chewed through buried gas, water, and electric lines. Gophers typically dig between 6 and 12 inches below the surface and are a particular threat to buried drops.

There are several ways to protect against gophers. The best protection is to bury fiber deep enough to be out of gopher range, but that can add a lot of cost to buried drops. I have a few clients that bore drops rather than trench or vibrate them for this reason. Another protection is to enclose the fiber in a sheathe that is over 3 inches in diameter. A tube of that size is too big for the gophers to bite. Again, this is an expensive solution for buried drops. Another solution is to surround the buried fiber with 6 – 8 inches of gravel of at least 1-inch size – anything smaller gets pushed to the side by gophers.

There are examples of even more exotic animal damage to fiber. Large birds of prey have sharp talons that can create small cuts in the sheathe and allow in water. Flocks of birds repeatedly sitting on a fiber can cause sag and stretching of the fiber. I can remember living in Florida and seeing end-to-end birds sitting on wires – that has to add a lot of weight to a 200-foot fiber between poles. Over the last year I’ve seen several reports of sharks chewing on undersea fibers.

Finally, although not directly animal related, a common cause of rural fiber cuts happens when farmers hit fiber with a backhoe when burying dead livestock. They typically bury wherever an animal died, including in the buried fiber right-of-way.

No Break in Global Chip Shortages

It looks like the chip shortage is going to last longer than expected. At the beginning of this year, it looked like the chip shortage might be over by early to mid-2027. That no longer seems to be the expectation.

As a reminder of the issue, the chip shortage has come about due to chip manufacturers migrating to produce the more profitable chips used by AI data centers. Recent predictions are that 70% of all memory chips manufactured worldwide will go to data centers for the rest of 2026 into 2027. To put that number into perspective, the first public AI model was introduced to the world in November 2022, and the demand for AI chips has grown since then. At that time, about 32% of chips went to the more traditional data centers used for cloud computing.

Electronics of all types rely on chips known as DRAM (Dynamic Random-Access Memory). A DRAM is comprised of a large number of tiny cells that contain one capacitor and one resistor. The simplicity of the structure makes it relatively easy to affordably manufacture chips in volume. The function of DRAM is to temporarily hold data in storage as the computer uses it, and to dynamically cycle in new data as needed during processing.

Broadband devices like Wi-Fi routers, cable modems, and the core electronics used for fiber and wireless networks use a type of DRAM called DDR4. A few other industries, like automotive and consumer electronics, also rely on DDR4 chips. DDR4 chips have less capacity and are cheaper to make than DDR5 chips used for computers and cellphones. By contrast, AI datacenters use HBM (high-bandwidth memory) chips. These are comprised of a three-dimensional stack of DRAM chips. HBM chips are larger and require more wafer space than normal DRAM, and the manufacturing process is more challenging. At least currently, there is a lot more profit in manufacturing the HBM chips.

Three companies have historically manufactured over 90% of all DRAM chips – SK Hynix Inc., Samsung Electronics Co., Ltd., and Micron Technology, Inc. All three companies have shifted a lot of manufacturing capacity to the HBM chips for AI. Micron fully halted the manufacture of DRAM for personal computers. SK Hynix converted its primary DRAM production line to AI chips in late 2024. DRAM wholesalers are alarmed that the average worldwide DRAM inventory in the supply chain has fallen from a normal 17-week supply in 2024 to a 2-4 week supply currently. All three companies have already pre-sold all of their DRAM capacity for the rest of 2026 and well into 2027.

Unfortunately, there is no short-term relief on the way. Micron is building a new mega-factory in Onondaga, New York, to manufacture DRAM that won’t benefit the supply chain until sometime in 2028. The company is expanding a factory in Taiwan that won’t produce new chips until at least the end of 2027. SK Hynix is investing $13 billion in a new factory in Cheongiu, South Korea, that will be completed at the end of 2027. Samsung is building as new factory in Gyeonngi Province in South Korea that won’t be completed until sometime in 2028.

It doesn’t seem realistic that any other companies will step in to fill the market void because of the complexity and intense quality control needed to make DRAM chips. This means a growing shortage of chips for telecom and other industries for the next two years. A few auto factories have already reduced the number of cars being built due to a lack of chips. For telecom, this is going to result in higher prices and longer waits to get electronics. This also means that ISPs at the bottom of the supply chain, those without the buying power of AT&T or Charter, will see the biggest impacts. Also note that ISPs are going to lose broadband customers who can’t afford to replace a dying computer. As always happens with a big price increase, I can’t imagine that chips will ever return to the old prices. The handful of manufacturers are going to expect higher profits from DRAM than in the past, and due to their monopoly, will be able to charge whatever they want.

Network Challenges from AI Traffic

Cisco recently published a report that looks at the impact of AI traffic on networks. It’s an interesting paper because Cisco found that AI traffic does not operate the same as most other web traffic. While the volume of AI traffic is small today, Cisco predicts that we’ll have to make changes to the web over time to accommodate growing AI traffic volumes. Cisco predicts that by 2035, one fourth of all web traffic will be AI agents and AI models in data centers.

Cisco notes that we’ve spent decades optimizing a web that delivers burst traffic, like video. When a video is viewed from the web, the data stream doesn’t have to be delivered evenly in real time. Instead, all that is needed is for the transmission of the video to reach the viewer before they are ready to watch it. Anybody who has watched a video can see that the streamed video is always working to to stay ahead of what you are watching.

AI traffic is very different. Cisco uses the term AI inference traffic to mean the real-time transfer of AI traffic between the AI models operating in data centers and users. AI inference traffic is delivered at what Cisco calls software speed, meaning the receiving end is ready to digest and use the data as it is delivered, quite different than streamed video that is only trying to stay ahead of a viewer.

The difference between AI traffic and normal web traffic is significant. The typical burst of AI traffic lasts twice as a typical burst of video data. While individual bursts of video data are smaller, the flow rate for video, which means the actual delivery time, lasts ten times longer than AI traffic, since video bursts are spread over time in multiple small bursts.

AI traffic is also two-way and requires a good upstream connection. In fact, Cisco found that 9% of AI traffic requires more upstream traffic than downstream traffic. Cisco believes the need for network upload speeds will increase as AI agents mature.

Current network latency is not a bottleneck for AI traffic, but Cisco says latency will become a problem as the volume of AI traffic increases. This will require a major rework of web architecture when latency becomes an issue.

Cisco found that tasks performed by AI generate 450% more traffic than the same task performed in a more traditional way. In Cisco’s vocabulary, AI agents act as power users and use a lot of network resources.

The bottom line is that AI traffic is different from current web traffic and will not only increase traffic volumes on networks, but it will also change the shape, symmetry, and needed priority of traffic.

There have already been discussions of creating a private web to connect between AI data centers. But that would only solve part of the problem, because AI traffic is eventually delivered to users throughout the web. AI traffic is going to create an interesting new set of challenges for network engineers, something that nobody envisioned just a few years ago.

Is There an AI Divide?

I recently attended and spoke at an AI conference. One of the things that became clear to me is that we are probably headed for a new digital divide related to AI. What do I mean by that?

In that short period of time, AI has touched a large majority of computer users. A survey from Pew in September 2025 showed that 95% of adults had heard of AI. At the time of the survey, 47% of people had heard a lot about AI, up from 26% measured in a 2022 survey. That’s bound to be a higher number in the summer of 2026. As you might expect, people with a high awareness of AI tend to be younger and better educated. For example, 62% of adults under 30 had heard a lot about AI, compared to 32% of those 65 and older. 60% of adults with post-graduate degrees had heard a lot about AI, versus 38% of those with a high school diploma or less.

Only about 3% of U.S. households pay for a consumer AI subscription. There is a lot of expectation that AI companies will be forced to greatly increase the costs of monthly access, which will undoubtedly lower the percentage of folks who are comfortable working directly with AI. Most people who use AI today interface through some tool like virtual assistants, GPS navigation apps, streaming algorithms, or weather forecasts. According to a survey from Quinnipiac University from March 2026, 51% of respondents say they use an AI tool, up from 37% a year earlier.

Another Pew poll showed that 46% of Americans hold a negative view of AI, while only 26% view it favorably. 57% believe the risks outweigh the benefits, and 41% actively distrust AI systems.

All of these statistics point to a possible AI-divide, separating those who believe in and use AI, from those who don’t trust AI and refuse to use it. This divide already exists, and the future question that will  have to be answered is if those who distrust AI will eventually be won over. It’s easy to forget that AI has only been available to the public since the end of 2022, which, for a new technology, is still in the infancy stages.

There will be consequences if there is a permanent gap between those who use AI and those who don’t or won’t. If AI brings measurable advantages to those who use it, then, over time, non-users will be at a disadvantage in many ways. There is also the possibility that those who distrust AI will be able to thwart its expansion. There are already numerous communities creating barriers to new data centers.

It’s easy to conclude that using AI is a choice, and that anybody who doesn’t use AI has to live with that decision. But consider a few statistics from the National Center for Education Statistics (NCES). It reports that 21% of adults are functionally illiterate, meaning they cannot complete basic reading tasks, such as filling out job applications, reading medicine instructions, or helping with children’s homework. 54% of adults have reading skills below a sixth-grade reading level. Less than half of adults demonstrate strong reading comprehension skills.

These statistics are relevant when talking about using and benefitting from AI, which is ultimately a language model. Literacy is far less of an issue when talking about the use of broadband, because it doesn’t require advanced reading skills to watch videos and engage in other forms of online entertainment and social media.

It’s way too early in the life of AI to draw any conclusions about a possible divide, but it’s not hard to foresee a likely divide between those who benefit from AI and those who can’t or won’t. In a decade, this might become the most important digital dividing line, more so than the digital skills divide we talk about today.