The largest radar ever built does not sit on a mountain. It floats. Six diesel generators, two hulls, a crew of somewhere between 48 and 86 depending on which part of the US government you ask, and on top of it a white sphere 36.5 metres (120 ft) across held up by nothing more sophisticated than the air pressure inside it. Under that dome is an X-band phased array so sensitive that the Missile Defense Agency once claimed it could pick a baseball out of the sky over San Francisco from Chesapeake Bay.

It is called the Sea-Based X-Band Radar, SBX-1, and it holds a Guinness World Record. It is also, by a considerable margin, the least useful of the four radars in this article.

That is the thing about the giants of the radar world. The ones with the records are not the ones doing the work. A 50-year-old array bolted to a rock in the Aleutians has spent half a century watching Soviet and then Russian missiles fall into the Pacific, and it is still meeting its availability requirement. The record holder has been in limited test support status since 2013. Size and usefulness point in opposite directions, and the gap between them is the story.

The record holder: a radar on an oil rig

The platform underneath SBX-1 began life as a fifth-generation CS-50 twin-hulled semi-submersible, bought from the Norwegian firm Moss Maritime in January 2003. It was converted at Keppel AmFELS in Brownsville, Texas, fitted with its radar mount at Kiewit Offshore in Ingleside, and named Sea-Based X-Band Radar-1 in July 2005. It went round the bottom of South America through the Strait of Magellan over the turn of 2005 into 2006. It moves under its own power at eight knots, driven by electric thrusters fed from six 3.6 MW Caterpillar diesel generators, and carries 1.8 million gallons of marine diesel.

The numbers on the radar itself need care, because the two US government bodies that publish them do not agree. The Missile Defense Agency gives 390 feet long, 240 feet wide, more than 280 feet from keel to the top of the dome, and a displacement of nearly 50,000 tons when ballasted. Military Sealift Command, which actually operates the thing, lists 389 feet, 238 feet, and 32,690 tons. MDA puts the crew at about 86; MSC says 48 civilian mariners, 27 of them in reduced operating status. Both are official. Neither has corrected the other.

The array is the part worth arguing about. MDA’s own history booklet says "approximately 45,000" transmit/receive modules. The figure of 22,000 that circulates widely comes from independent accounts describing an array built for roughly 45,000 positions but populated with about half that number, concentrated toward the centre to suppress grating lobes. Both statements can be true at once: the structure has 45,000 slots and rather fewer modules in them. MDA has never published the lower number.

Guinness World Records adjudicated the record in 2006 under the exact title "Largest radar, phased array X-band". It is worth quoting the record and not the reasoning, because the Guinness entry also states that the radar "operates within the X-ray band of the electromagnetic spectrum". X-band is microwave, roughly 8 to 12 GHz. X-rays are about nine orders of magnitude up from there. The record stands; the physics in the citation does not.

The Missile Defense Agency’s own film on the SBX, including footage from inside the radome.

The soda straw problem

The criticism of SBX has never been that it is not sensitive. It is that it cannot look at very much at once. The instantaneous electronic scan field of view is roughly 25 degrees, against 90 to 120 degrees for a conventional radar of comparable class. The array and platform can be physically turned, which is where the much wider figures quoted elsewhere come from, but turning a 50,000-ton oil rig is not the same manoeuvre as steering a beam.

Then there is the environment. The Project On Government Oversight published a detailed critique in December 2006 listing electronics unsuited to saltwater, ice loading risk, an absence of dual-redundant electronics, no EMP hardening and no secure fibre link. The same piece quoted the plainest version of the objection.

Senator Ted Stevens of Alaska put the same worry more economically to Lieutenant General Ronald Kadish, then director of the Missile Defense Agency, during a Senate hearing on the wisdom of parking a sensor of that value in the North Pacific: “I hope your people are nautical enough to know what you’re doing.”

Cost figures vary by more than a factor of two depending on what is being counted. Guinness records $900 million to develop. POGO reported $1 billion "and counting" by the end of 2006. Reporting in the Los Angeles Times put the total at $2.2 billion, a figure the Honolulu Star-Advertiser headline rendered as a "$2.2 billion boondoggle".

On paper SBX-1 is home-ported at Adak, Alaska, where a $26 million mooring was installed in 2007. In practice it lives at Ford Island, Joint Base Pearl Harbor-Hickam. In November 2025 its radome — the original, after twenty years — was deflated and removed for the first time. The replacement arrived in a 143-foot container, was lifted into place by crane with 200 line-handlers positioning it, and was inflated on 10 December 2025.

The full maintenance availability ran about two and a half months and finished in February 2026. So the largest radar ever built is maintained, funded and seaworthy. It is also, by the Government Accountability Office’s account in June 2023, the only one of nine missile defence elements reviewed that had no approved sustainment plan.

Cobra Dane: fifty years on a rock in the Aleutians

Shemya is a treeless island about 1,500 km from the Alaskan mainland and roughly 500 km from Kamchatka. That second number is the entire point. Cobra Dane was built by Raytheon in 1976 and turned over to Aerospace Defense Command in 1977 — the Federation of American Scientists says July 1977, the GAO says it became operational in 1976, and the disagreement has never been resolved — to watch Soviet intercontinental and submarine-launched ballistic missiles arrive in the Kamchatka impact area. It was built to verify SALT II.

The face is 29 metres (95 ft) across, with 34,768 element positions of which 15,360 are active, space-tapered down to 20 per cent density at the edges. That element count is the best-sourced figure in this article: it traces back through George N. Lewis to the original engineering literature from 1976 and 2000, rather than to a fact sheet. It works in L-band with a peak power of 15.4 MW, average 920 kW, a 0.6-degree beamwidth and a range that reached about 14,000 km as of 2000, sufficient to detect objects of 4 to 5 cm.

Half a century later it is doing a different job. It is a midcourse sensor for ballistic missile defence and a space surveillance asset, and the Space Force assumed authority for it from December 2019. The GAO reported in December 2018 that it was meeting its operational availability requirement despite key components being obsolete, and that when Cobra Dane is down, the backup is SBX. The older radar covers for the bigger one.

Keeping it alive is not free. The Air Force and MDA planned $278.6 million for operation and sustainment plus roughly $140 million for Shemya site maintenance. V2X took a $62 million contract in April 2025 running to March 2027. The GAO’s 2023 review named two specific, almost domestic problems: HVAC systems that only partly cool the radar, and the fact that Pacific Air Forces runs Eareckson Air Station and "does not always prioritize Cobra Dane-specific issues highly". A radar that watches for nuclear attack can be undermined by its own air conditioning.

Cobra Dane is not Cobra Ball

This confuses people constantly, and it should, because the names were never meant to describe hardware. "Cobra" is a Defense Intelligence programme designator. Cobra Dane is the fixed array on Shemya. Cobra Ball is three RC-135S aircraft flown by the 55th Wing out of Offutt AFB, Nebraska, carrying optical and electronic sensors. Cobra Judy was a shipborne radar on USNS Observation Island, now retired and replaced by Cobra King aboard USNS Howard O. Lorenzen. Cobra Mist was an over-the-horizon radar at Orford Ness in Suffolk that never worked and was shut down on 30 June 1973.

One is a 29-metre array bolted to a rock in the Aleutians. Another is a converted airliner out of Nebraska. They are siblings by mission and funding lineage, not by engineering. And while we are separating names: PAVE PAWS has nothing to do with the HH-60G Pave Hawk helicopter or with Paveway guided bombs either.

PAVE PAWS: the triangles, and an official error

PAWS is Phased Array Warning System, on which every source agrees. PAVE is an Air Force programme code and its expansion is genuinely disputed — GlobalSecurity states flatly that it does not have one, other accounts render it "Precision Acquisition Vehicle Entry", and the US Space Force fact sheet simply declines to expand it. Treat it as a code word that may have acquired a backronym.

The geometry is the memorable part. Two array faces are set into the sloping sides of a three-sided building 32 metres (105 ft) high, oriented 120 degrees apart. Each face steers 60 degrees either side of boresight. That gives 120 degrees per face and 240 degrees of azimuth in total. Each face is 31 m (102 ft) wide with a 22.1 m active aperture, carries 2,677 crossed-dipole element positions of which 1,792 are active, runs in UHF between 420 and 450 MHz with circular polarisation and 580 kW peak power per array, has a 2.2-degree beamwidth, and reaches about 3,000 nautical miles.

Here is a detail worth keeping. The official US Space Force fact sheet for the system states that the two faces each cover "360 degrees azimuth". That is geometrically impossible for a fixed planar array — a flat face cannot see behind itself, and two of them cannot each see everywhere. Every technical source gives 120 per face and 240 total. The error sits in the primary source, uncorrected, which is a useful reminder that "official" and "correct" are different categories.

Four were built originally: Otis AFB on Cape Cod and Beale AFB in California in 1980–81, then Robins AFB in Georgia in 1986 and Eldorado AFS in Texas in 1987. The latter two closed in 1995. Today five Upgraded Early Warning Radars, AN/FPS-132, are active: Cape Cod Space Force Station with the 6th Space Warning Squadron, Beale with the 7th, Clear in Alaska with the 13th, Pituffik Space Base in Greenland — formerly Thule Air Base, renamed on 6 April 2023 — with the 12th, and RAF Fylingdales in Yorkshire under UK Space Command.

JORN: the one that solved the problem that killed the others

Australia’s Jindalee Operational Radar Network is the odd one out here, because it is not a single structure at all. Three radars, each bistatic with transmit and receive sites about 100 km apart: Longreach in Queensland transmitting with its receiver at Stonehenge, Laverton in Western Australia transmitting from Leonora, and Alice Springs in the Northern Territory transmitting from Harts Range and receiving at Mount Everard. It is run by No. 1 Remote Sensor Unit, feeding the JORN Coordination Centre at RAAF Base Edinburgh.

Australian Defence Science and Technology gives the official range as wide-area surveillance at 1,000 to 3,000 kilometres, working between 5 and 30 MHz. Alice Springs and Longreach cover 90 degrees each, Laverton 180. You will see coverage-area figures in square kilometres quoted elsewhere; the commonly repeated one describes a box smaller than Tasmania for a radar that sees 3,000 km, so we are printing the range and leaving the area alone.

Australia’s Defence Science and Technology Group on how JORN bounces signal off the ionosphere.

Over-the-horizon radar has a graveyard attached to it. Cobra Mist, the Anglo-American AN/FPS-95 at Orford Ness, cost $100–150 million and was abandoned in 1973 after testing found unexplained noise that made it report missile launches no matter where it was pointed. Internal equipment faults were eliminated as the cause. The distortion only appeared over land. An SRI International panel spent from January to May 1973 on it and no convincing explanation was ever found. The Soviet Duga array — the Russian Woodpecker — has its own long story, which we told separately.

What JORN does differently is measure the thing everyone else assumed. The network runs twelve vertical ionosondes, Single-Receiver Digisonde Portable Sounder units built by the University of Massachusetts Lowell, generating a fresh ionospheric map every 225 seconds. The ionosphere is not treated as a mirror that is simply there. It is treated as a medium whose behaviour has to be re-measured continuously, across a continent, because it changes. Nobody has published a causal claim that this is why JORN works and Cobra Mist did not, so we will not make one. But the contrast in the two records is hard to miss.

Sinnott also offers the most human line in the whole radar literature, on the standard he and his colleagues held themselves to: “I think, in retrospect, we probably could have got away with a little lesser performance, but we young guys were going to make this as good as we could.”

It is not cheap and it is not on time. The Phase 6 upgrade, announced in March 2018 with BAE Systems Australia as prime, has a total approved budget of A$1,250.4 million, of which A$538.8 million had been spent by 30 June 2025. Its Preliminary Design Review was planned for October 2019 and achieved in September 2023 — a variance of 47 months. The initial materiel release, IOC and FOC dates are marked "Not For Publication", which means withheld rather than unknown. Defence states there is sufficient budget remaining to complete the agreed scope.

And it works well enough that other countries want it. Australia and Canada signed an over-the-horizon radar technology partnership on 18 July 2025, covering Arctic OTHR challenges and Canadian development.

No, JORN did not see MH370

This one recurs every few years, so it is worth settling. The Australian Department of Defence has addressed it directly: based on the time of day the aircraft disappeared, and in the context of peacetime tasking, JORN was not operational at the time. Defence adds that given the range from the individual radars, the ionospheric conditions and the absence of information about MH370’s possible path toward Australia, it is unlikely the aircraft would have been detected even if the system had been running. Every claim to the contrary traces back to conspiracy blogs rather than to Defence.

Eight old radars, one digital rebuild

The most consequential radar story of 2026 is not a new structure. It is a software and signal-processing programme. In May 2026 the Space Force confirmed Ground Based Radar Digitization: eight legacy radars converted from analogue to digital. The five Upgraded Early Warning Radars, Cobra Dane, PARCS at Cavalier Space Force Station in North Dakota, and the C-6 at Eglin AFB in Florida. Vendor selection was set for June 2026, preliminary design review for 2027, with rapid prototyping at up to four sites and the remaining four beginning rapid fielding in FY2030. The budget is $128 million in FY2027 and $654 million across FY2028 to 2031. The systems are officially described as facing critical obsolescence and sustainment challenges.

Which brings the whole thing round to where it started. The radar with the world record spent the winter of 2025–26 in a shipyard having its balloon replaced, and has been in limited test support status for thirteen years. The radars getting the money are a 1977 array on an Aleutian rock, four Cold War triangles and a couple of others, none of which would win a superlative for anything. Nobody is filming a documentary about a transmitter group replacement on Shemya. It is simply the part that has to keep working.

SBX-1 under tow into Pearl Harbor. For scale, the dome is as tall as a ten-storey building.

Sources: Missile Defense Agency SBX fact sheet and history booklet; Military Sealift Command ship inventory; Guinness World Records; DVIDS (MSC, 17 December 2025); Hawaii News Now; Project On Government Oversight (December 2006); GAO-19-68 and GAO-23-105578; Federation of American Scientists; George N. Lewis, mostlymissiledefense.com, citing Chorman (2000) and Filer & Hartt (1976); US Space Force fact sheets; Air & Space Forces Magazine (11 May 2026); airforce-technology.com; Australian Defence Science and Technology Group; Australian National Audit Office PDSS AIR2025 Phase 6 (2024-25 Major Projects Report); Australian Department of Defence JORN fact sheet; GlobalSecurity.org.